Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberCA2143672 C
Publication typeGrant
Application numberCA 2143672
PCT numberPCT/US1994/007397
Publication date1 Feb 2005
Filing date8 Jul 1994
Priority date8 Jul 1993
Also published asCA2143672A1, EP0634719A2, EP0634719A3, US5603031, US6016393, WO1995002219A1
Publication numberCA 2143672, CA 2143672 C, CA 2143672C, CA-C-2143672, CA2143672 C, CA2143672C, PCT/1994/7397, PCT/US/1994/007397, PCT/US/1994/07397, PCT/US/94/007397, PCT/US/94/07397, PCT/US1994/007397, PCT/US1994/07397, PCT/US1994007397, PCT/US199407397, PCT/US94/007397, PCT/US94/07397, PCT/US94007397, PCT/US9407397
InventorsJames E. White, Christopher S. Helgeson, Douglas A. Steedman
ApplicantJames E. White, Christopher S. Helgeson, Douglas A. Steedman, General Magic, Inc., Intellectual Ventures Patent Holding I, Llc, Ben Franklin Patent Holding Llc
Export CitationBiBTeX, EndNote, RefMan
External Links: CIPO, Espacenet
System and method for distributed computation based upon movement, execution and interaction of processes in a network
CA 2143672 C
Abstract
A distributed computing environment in which agent processes (150A) direct their own movement through a computer network (100).
Place processes (220A, 220B) provide a computing context within which agent processes (150A) are interpreted. An agent process (150A) controls its movement from one place process (220A) to another (220B) within the network (100) by using a ticket (1306). An agent process (150A) which moves from one place process (220A) to another (220B) transports definitions of classes of which objects included in the agent process (150A) are members. An agent process (150A) which moves from one place process (220A) to a second place process (220B) avoids unnecessary transportation of objects included in the agent process (150A) by substituting equivalent objects which are found in the second place process (220B).
Claims(165)
1. A method for implementing remote programming in a computer network comprising the steps of:
defining a plurality of object-oriented classes including an agent class and a place class;
forming instructions for a computer process, said instructions including said object-oriented classes, subclasses of said object-oriented classes, and a go operation; and interpreting said instructions on a processor in said computer network wherein, in response to said go operation, an agent process is transported to a place process and further wherein said agent process is a member of said agent class and said place process is a member of said place class.
2. The method for implementing remote programming in a computer network as in Claim 1 further comprising forming said go operation within said agent process.
3. The method for implementing remote programming in a computer network as in Claim 1 further comprising forming a subplace of said place wherein said subplace is a member of said place class and said place is a superplace of said subplace.
4. The method for implementing remote programming in a computer network as in Claim 1 further comprising designating said agent process as an owner of an object.
5. The method for implementing remote programming in a computer network as in Claim 4 further comprising forming within said object a digest.
6. The method for implementing remote programming in a computer network as in Claim 5 further comprising interchanging for said first-mentioned object a second object at said place process wherein said second object has a second digest equivalent to said first-mentioned digest and further wherein said first object is not transported to said place process.
7. The method for implementing remote programming in a computer network as in Claim 1 further comprising specifying said place process by a ticket means.
8. The method for implementing remote programming in a computer network as in Claim 7 further comprising forming said ticket means within said agent process.
9. The method for implementing remote programming in a computer network as in Claim 7 further comprising forming within said ticket means an address of said place process.
10. The method for implementing remote programming in a computer network as in Claim 9 wherein said address is a teleaddress.
11. The method for implementing remote programming in a computer network as in Claim 7 further comprising forming within said ticket means a name of said place process.
12. The method for implementing remote programming in a computer network as in Claim 11 wherein said name is a telename.
13. The method for implementing remote programming in a computer network as in Claim 1 further comprising:

defining a class of class objects;
forming a class object, wherein said class object (i) is a member of said class of class objects, and (ii) defines a first class which is a subclass of a selected one of said object-oriented classes; and forming an object, said object being a member of said first class;
wherein said agent process owns said object; and transportation of said agent process to said place process comprises transporting said object and said class object to said place process.
14. A method for implementing remote programming in a computer network comprising the steps of:
defining a plurality of object-oriented classes including an agent class and a place class;
forming instructions for a computer process including said object-oriented classes, subclasses of said object-oriented classes, and a send operation;
and interpreting said instructions on a processor in said computer network wherein, in response to said send operation, a clone of an agent process is transported to a place process and further wherein said clone and said agent process are each a member of said agent class and said place process is a member of said place class.
15. The method for implementing remote programming in a computer network as in Claim 14 further comprising forming said send operation within said agent process.
16. The method for implementing remote programming in a computer network as in Claim 14 further comprising forming a subplace of said place wherein said subplace is a member of said place class and said place is a superplace of said subplace.
17. The method for implementing remote programming in a computer network as in Claim 14 further comprising designating said agent process as an owner of an object.
18. The method for implementing remote programming in a computer network as in Claim 17 wherein transportation of said clone comprises transporting a copy of said object to said place process.
19. The method for implementing remote programming in a computer network as in Claim 18 further comprising forming within said object a digest.
20. The method for implementing remote programming in a computer network as in Claim 19 further comprising interchanging for said copy a second object, which is different from said first-mentioned object, at said place process wherein said second object has a second digest equivalent to said first-mentioned digest and further wherein said copy is not transported to said place process.
21. The method for implementing remote programming in a computer network as in Claim 14 further comprising specifying said place process by a ticket means.
22. The method for implementing remote programming in a computer network as in Claim 21 further comprising forming said ticket means within said agent process.
23. The method for implementing remote programming in a computer network as in Claim 21 further comprising forming within said ticket means an address of said place process.
24. The method for implementing remote programming in a computer network as in Claim 23 wherein said address is a teleaddress.
25. The method for implementing remote programming in a computer network as in Claim 21 further comprising forming within said ticket means a name of said place process.
26. The method for implementing remote programming in a computer network as in Claim 25 wherein said name is a telename.
27. The method for implementing remote programming in a computer network as in Claim 14 further wherein, in response to said send operation, a second clone of said agent process is transported to a second place process wherein said second clone is a member of said agent class and is distinct from said first-mentioned clone and further wherein said second place process is a member of said place class.
28. The method for implementing remote programming in a computer network as in Claim 27 further comprising:
determining that transportation of said first clone to said first place process and of said second clone to said second place process requires transportation of said first and second clones to a single computer of a plurality of computers of said computer network;
transporting said first clone to said single computer; and forming, in said single computer, said second clone from said first clone.
29. The method for implementing remote programming in a computer network as in Claim 28 further comprising:
transporting said first clone from said single computer to said first-mentioned place process; and transporting said second clone from said single computer to said second place process.
30. A method for interprocess communication in a computer comprising the steps of:
defining a plurality of object-oriented classes including an agent class;
forming instructions for a computer process including said object-oriented classes, subclasses of said object-oriented classes, and a meet operation;
and interpreting said instructions on a processor in said computer network wherein, in response to said meet operation, a meeting place process provides a first agent process access to a second agent process and provides said second agent process access to said first agent process and further wherein said first and second agent processes are members of said agent class.
31. The method for interprocess communication in a computer as in Claim 30 wherein said meeting place process is a member of a place class in said plurality of object-oriented classes and further wherein said first and second agent processes occupy said meeting place process.
32. The method for interprocess communication in a computer as in Claim 31 further comprising forming a second place process wherein:

said second place process is a member of said place class;
said second place process is a subplace of said meeting place process; and said meeting place is a superplace of said second place.
33. A method for controlling movement of processes in a computer network, said method comprising the steps of:
defining a plurality of object-oriented classes including an agent class, a place class, and a ticket class;
forming a plurality of place processes in said computer network wherein each of said plurality of place process is a member of said place class;
forming an agent process wherein said agent process is a member of said agent class and occupies a first place process in said plurality of place processes; and forming a ticket wherein said ticket is a member of said ticket class and defines a trip involving the movement of said agent process from said first place process to a second place process in said plurality of place processes.
34. A method for limiting capabilities of processes in a computer network, said method comprising:
defining a plurality of object-oriented classes including a process class and a permit class;
forming a process wherein said process is a member of said process class; and forming a permit wherein said permit is a member of said permit class and specifies one or more capabilities of said process.
35. The method as in Claim 34 further comprising:
defining within said process class a go operation; and specifying within said permit whether said process is capable of performing said go operation.
36. The method as in Claim 34 further comprising:
defining within said process class a send operation; and specifying within said permit whether said process is capable of performing said send operation.
37. The method as in Claim 34 further comprising:
defining within said process class a charge operation; and specifying within said permit whether said process is capable of performing said charge operation.
38. The method as in Claim 34 further comprising:
defining within said process class a terminate operation; and specifying within said permit whether said process is capable of performing said terminate operation.
39. The method as in Claim 34 further comprising specifying within said permit whether said process is capable of creating a second process, different from said first-mentioned process, wherein said second process is a member of said process class.
40. The method as in Claim 34 further comprising:
defining an object-oriented place class; and specifying within said permit whether said process is capable of creating members of said place class.
41. The method as in Claim 34 further comprising specifying within said permit whether said process is restarted upon failure of said process.
42. The method as in Claim 34 further comprising specifying within said permit an amount of processing that is allotted to said process.
43. The method as in Claim 34 further comprising specifying within said permit a time at which said process is terminated.
44. The method as in Claim 34 further comprising:
defining within said process class a restrict operation;
forming instructions for a computer process, said instructions including said classes, subclasses of said classes, and said restrict operation;
interpreting said instructions on a processor in said computer network wherein, in response to said restrict operation, a second permit, which is different from said first-mentioned permit, is formed and further wherein said second permit is a member of said permit class and specifies a group of one or more capabilities of said one or more capabilities of said process; and restricting said process to said group of one or more capabilities specified by said second permit.
45. In a computer, a method of interpreting processes of various versions of an instruction set, said method comprising:

defining a plurality of object-oriented classes including a class of classes and a class of citations;
forming one or more class objects wherein said class objects are members of said class of classes;
forming within a first of said class objects a citation wherein said citation specifies said first class object and specifies which of said class objects are backward compatible with said first class object.
46. In a computer network having a plurality of computers, a communication process comprising:
providing a plurality of place processes within said computer network wherein each place process is a locale in one of said computers for zero or more agent processes;
specifying, by a ticket means, a trip for an agent process to a destination place process in said plurality of place processes; and transporting, in response to a send operation within said agent process, a clone of said agent process to said destination place process.
47. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, by a second ticket means, a second trip, which is different from said first-mentioned trip, for said agent process to a second destination place process in said plurality of place processes.
48. In a computer network having a plurality of computers, the communication process as in Claim 47 further comprising:
determining that said first-mentioned clone, in taking said first trip, and a second clone, in taking said second trip, are both to be transported to a single computer of said computers;
moving to said single computer said first clone;
and forming within said single computer said second clone from said first clone.
49. In a computer network having a plurality of computers, the communication process as in Claim 48 further comprising transporting said first clone to said first-mentioned destination place process.
50. In a computer network having a plurality of computers, the communication process as in Claim 48 further comprising transporting said second clone to said second destination place process.
51. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising transporting, in response to a go operation within said agent process, said agent process to said destination place process specified by said ticket means.
52. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying said destination place process by a name within said ticket means.
53. In a computer network having a plurality of computers, the communication process as in Claim 52 wherein said name is a telename.
54. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying said destination place process by an address within said ticket means.
55. In a computer network having a plurality of computers, the communication process as in Claim 54 wherein said address is a teleaddress.
56. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, within said ticket means, said destination place process by a citation of a class of which said destination place process is a member.
57. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, within said ticket means, said destination place process by any combination of an address of said destination place process, a name of said destination place process, and a citation of a class of which said destination place process is a member.
58. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, within said ticket means, a maximum time period for said trip.
59. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, within said ticket means, a desired time period for said trip.
60. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, within said ticket means, a transportation means wherein said transportation means specifies a type of intercomputer communications means by which said clone is transported.
61. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising controlling, with a destination permit contained in said ticket means, a deadline for said agent process at said destination place process.
62. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising controlling, with a destination permit contained in said ticket means, operations that said agent process is allowed to perform at said destination place process.
63. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising controlling, with a destination permit contained in said ticket means, resources that said agent process is allowed to consume at said destination place process.
64. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising specifying, with a destination permit contained in said ticket means, a priority for said agent process at said destination place process relative to other processes at said destination place process.
65. In a computer network having a plurality of computers, the communication process as in Claim 46 further comprising controlling, via a permit means, a capability that a process is allowed to have.
66. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said permit means is a native permit.
67. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said permit means is a local permit.
68. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said permit means is a temporary permit.
69. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said capability comprises resources that said process can consume.
70. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said capability comprises a deadline after which the process can not proceed.
71. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said capability comprises a priority for said process relative to other processes.
72. In a computer network having a plurality of computers, the communication process as in Claim 65 wherein said capability comprises permission for said process to perform selected operations.
73. In a computer network having a plurality of computers, the communication process as in Claim 46 wherein said agent process owns an object.
74. In a computer network having a plurality of computers, the communication process as is Claim 73 wherein said object has a digest.
75. In a computer network having a plurality of computers, the communication process as is Claim 74 wherein said step of transporting a clone of said agent process comprises transporting a copy of said object.
76. In a computer network having a plurality of computers, the communication process as is Claim 75 further comprising interchanging for said copy a second object at said destination place process wherein said second object has a digest equivalent to said first-mentioned digest and further wherein said copy is not transported to said destination place process.
77. In a computer network having a plurality of computers, the communication process as in Claim 73 further comprising the step using an interchanged object for said object at said destination place process thereby eliminating the need to transport said object to said destination place process.
78. In a computer network having a plurality of computers, the communication process as in Claim 46 wherein said transporting step further comprises a step of entering said destination place process.
79. In a computer network having a plurality of computers, the communication process as in Claim 46 wherein said transporting step further comprises a step of exiting said source place process.
80. In a computer network having a plurality of computers, the communication process as in Claim 46 wherein said destination place process is a meeting place process and said meeting place arranges a meeting between a requestor agent process and a petitioned agent process.
81. In a computer, a communication process comprising:
providing a first agent process and a second agent process;
specifying a meeting between said first and second agent processes by a petition means; and arranging said meeting between said first and second agent processes as defined by said petition means.
82. In a computer, the communication process as in Claim 81 further comprising forming said petition means within said first agent process.
83. In a computer, the communication process as in Claim 81 wherein said step of specifying a meeting comprises specifying within said petition means said second agent process.
84. In a computer, the communication process as in Claim 83 wherein said step of specifying said second agent process comprises specifying, within said petition means, said second agent by a name.
85. In a computer, the communication process as in Claim 84 wherein said name is a telename.
86. In a computer, the communication process as in Claim 83 wherein said step of specifying said second agent process comprises specifying, within said petition means, a class of which said second agent process is a member.
87. In a computer, the communication process as in Claim 83 wherein said step of specifying said second agent process comprises specifying, within said petition means, a citation of a class of which said second agent process is a member.
88. In a computer, the communication process as in Claim 81 wherein said step of specifying a meeting comprises specifying, within said petition means, a maximum time period for arranging said meeting.
89. In a computer, the communication process as in Claim 81 wherein said step of arranging said meeting comprises providing to said first agent process a reference to said second agent process.
90. In a computer, the communication process as in Claim 89 wherein said step of arranging said meeting further comprises providing to said second agent process a reference to said first agent process.
91. In a computer, the communication process as in Claim 81 further comprising causing, in response to an instruction within said first agent process, performance of an operation within said second agent process.
92. In a computer, the communication process as in Claim 91 wherein said first agent process owns an object.
93. In a computer, the communication process as in Claim 92 further comprising providing to said second agent process a reference to said object.
94. In a computer, the communication process as in Claim 93 wherein said reference is a protected reference.
95. In a computer, the communication process as in Claim 92 further comprising providing to said second agent process a reference to a copy of said object.
96. In a computer, the communication process as in Claim 91 wherein said second agent process owns an object.
97. In a computer, the communication process as in Claim 96 further comprising providing to said first agent process a reference to said object.
98. In a computer, the communication process as in Claim 97 wherein said reference is a protected reference.
99. In a computer, the communication process as in Claim 96 further comprising providing to said first agent process a reference to a copy of said object.
100. In a computer network having a plurality of computers, a communication system comprising:
an agent means having a ticket means and a send operation; and a plurality of place means wherein each place means is operative in one of said plurality of computers;
wherein said agent means is at a first place means in said plurality of place means;
said ticket means specifies a trip for said agent means to a destination place means in said plurality of place means; and said send operation transports a clone of said agent means to said destination place means.
101. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said agent means further comprises a second ticket means, different from said first-mentioned ticket means, that specifies a second trip to a second destination place means for said agent means.
102. In a computer network having a plurality of computers, the communication system as in Claim 101 wherein performance of said send operation transports a second clone of said agent means to said second destination place means.
103. In a computer network having a plurality of computers, the communication system as in Claim 102 wherein:
said first-mentioned trip and said second trip both include transportation to a single computer in said plurality of computers; and further wherein, in transporting said first-mentioned clone to said first-mentioned destination place means and said second clone to said second destination place means, performance of said send operation:
(a) transports said first clone to said single computer and (b) forms said second clone from said first clone within said single computer.
104. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said agent means further comprises a go operation wherein said go operation transports said agent means to a third destination place means specified by a third ticket means.
105. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a name which identifies said destination place means.
106. In a computer network having a plurality of computers, the communication system as in Claim 105 wherein said name is a telename.
107. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises an address for said destination place means.
108. In a computer network having a plurality of computers, the communication system as in Claim 107 wherein said address is a teleaddress.
109. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a citation means which specifies a class of which said destination place means is a member.
110. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises any combination selecting from the group consisting of an address of said destination place means, a name of said destination place means, and a citation means which specifies a class of which said destination place means is a member.
111. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises means for specifying a maximum time period for said trip.
112. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises means for specifying a desired time period for said trip.
113. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a way means for specifying a type of intercomputer communications means by which said trip is to be accomplished.
114. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a permit means for controlling a deadline for said agent means at said destination place means.
115. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a permit means for controlling operations that said agent means is allowed to perform at said destination place means.
116. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a permit means for controlling resources that said agent means is allowed to consume at said destination place means.
117. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said ticket means further comprises a permit means for specifying a priority for said agent means at said destination place means relative to other agent means at said destination place means.
118. In a computer network having a plurality of computers, the communication system as in Claim 100 further comprising a permit means for controlling a capability that an agent means is allowed to have.
119. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said permit means is a native permit.
120. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said permit means is a local permit.
121. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said permit means is a temporary permit.
122. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said capability comprises resources that said agent means can consume.
123. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said capability comprises a deadline after which the agent means can not proceed.
124. In a computer network having a plurality of computers, the communication system as in Claim 118 further comprising other agent means wherein said capability comprises a priority for said agent means relative to said other agent means.
125. In a computer network having a plurality of computers, the communication system as in Claim 118 wherein said capability comprises permission for said agent means to perform selected operations.
126. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said agent means owns an object.
127. In a computer network having a plurality of computers, the communication system as is Claim 126 wherein said object has a digest.
128. In a computer network having a plurality of computers, the communication system as is Claim 127 wherein a copy of said object is transported with said clone by performance of said send operation.
129. In a computer network having a plurality of computers, the communication system as is Claim 128 wherein a second object at said destination place means is interchanged for said copy; and further wherein said second object has a digest equivalent to said first-mentioned digest, thereby obviating transportation of said copy to said destination place means.
130. In a computer network having a plurality of computers, the communication system as in Claim 128 wherein said send operation uses an interchanged object for said copy at said destination place means thereby eliminating the need to transport said copy of said object to said destination place means.
131. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said place means further comprises means for entering said place means.
132. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said place means further comprises means for exiting said place means.
133. In a computer network having a plurality of computers, the communication system as in Claim 100 wherein said place means further comprises meeting place means wherein said meeting place means arranges a meeting between a requestor agent means and a responder agent means.
134. In a computer, a communication system comprising:
meeting place means having a meet operation wherein said meeting place is a locale for a plurality of agent means;
a petition means;
a first agent means in said plurality of agent means; and a second agent means in said plurality agent means wherein;
said first and second agent means can meet with any of said plurality of agent means;
said petition means specifies a meeting between said first and second agent means; and in response to said meet operation, said meeting place means arranges said meeting.
135. In a computer, the communication system as in Claim 134 wherein said first agent means contains said petition means.
136. In a computer, the communication system as in Claim 135 wherein said petition means specifies said meeting by specifying said second agent process.
137. In a computer, the communication system as in Claim 136 wherein said petition means comprises a name which specifies said second agent.
138. In a computer, the communication system as in Claim 137 wherein said name is a telename.
139. In a computer, the communication system as in Claim 136 wherein said petition means specifies said second agent by specifying a class of which said second agent process is a member.
140. In a computer, the communication system as in Claim 136 wherein said petition means comprises a citation, said citation specifying a class of which said second agent process is a member.
141. In a computer, the communication system as in Claim 134 wherein said petition means defines a maximum time period for arranging said meeting.
142. In a computer, the communication system as in Claim 134 further wherein said meet operation provides to said first agent means a reference to said second agent means.
143. In a computer, the communication system as in Claim 142 further wherein said meet operation provides to said second agent means a reference to said first agent means.
144. In a computer, the communication system as in Claim 134 further wherein said first agent means can cause performance of an operation within said second agent means.
145. In a computer, the communication system as in Claim 144 wherein said first agent means owns an object.
146. In a computer, the communication system as in Claim 145 further wherein said first agent means can provide to said second agent means a reference to said object.
147. In a computer, the communication system as in Claim 146 wherein said reference is a protected reference.
148. In a computer, the communication means as in Claim 145 further wherein said first agent means can provide to said second agent means a reference to a copy of said object.
149. In a computer, the communication system as in Claim 144 wherein said second agent means owns an object.
150. In a computer, the communication system as in Claim 149 further wherein said second agent means can provide to said first agent means a reference to said object.
151. In a computer, the communication system as in Claim 150 wherein said reference is a protected reference.
152. In a computer, the communication system as in Claim 149 further wherein said second agent means can provide to said first agent means a reference to a copy of said object.
153. In a computer network having one or more computers, a method for transferring data from a first engine process to a second engine process, said method comprising the steps of:
(a) providing means for executing one or more agent processes wherein said executing means comprises said first and second engine processes and further wherein each said agent process comprises instructions from a computer instruction set and has an execution state;
(b) providing within said computer instruction set a go instruction wherein said go instruction is contained within a first agent process executing within said first engine process and further wherein performance of said go instruction causes:
(i) suspension of execution of said first agent process by said first engine process;
(ii) representation of said first agent process such that said execution state of said first agent process is preserved;
(iii) transfer of said representation of said first agent process from said first engine process to said second engine process;
and (iv) resumption of execution of said first agent process by said second engine process;
(c) causing execution of said agent process by said first engine process, thereby causing performance of said go instruction.
154. The method of Claim 153 wherein said computer instruction set is object-oriented.
155. The method of Claim 154 wherein said plurality of agent processes are objects of said object-oriented computer instruction set.
156. The method of Claim 153 further comprising the step of adding data to said first agent process prior to causing performance of said go instruction;
wherein said representation of said first agent process includes said data; and further wherein transfer of said representation of said first agent process to said second engine process includes transfer of said data.
157. The method of Claim 153 wherein said first agent process comprises data representing a message to be transported from said first engine process to said second engine process; and further wherein transfer of said first agent process resulting from execution of said go instruction causes transfer of said data from said first engine process to said second engine process.
158. The method of Claim 153 wherein said first engine process executes on a first computer and said second engine process executes on a second computer wherein said first and second computers are part of said computer network.
159. A method for transferring data from a first engine process to one or more engine processes, said method comprising the steps of:
(a) providing means for executing a plurality of agent processes which comprise instructions from a computer instruction set wherein said agent process execution means comprises said first and one or more engine processes and further wherein each said agent process has an execution state;
(b) providing within said computer instruction set a send instruction wherein said send instruction is contained within a first agent process in said plurality of agent processes and further wherein, execution of said send instruction, includes the steps of:
(i) forming one or more copies of said first agent process wherein said copies preserve and include said execution state of said first agent process;
(ii) transferring each of said copies of said first agent process from said first engine process to a respective one of said one or more engine processes; and (iii) effectuating execution of each of said copies of said first agent process by said respective one of said one or more engine processes so as to simulate the resumed execution of said first agent process.
160. The method of Claim 159 wherein, upon a condition wherein two or more of said copies of said first agent process are to be transferred from said first engine process to a single second engine process, the transfer of said two or more copies of said first agent process comprises the steps of:

transferring to said second engine process a single copy of said first agent process; and forming within said second engine process from said single copy said two or more copies of said first agent process.
161. A method of transferring data from a first agent process which is executing within a computer system to a second agent process which is executing within said computer system wherein said first and second agent processes are occupants of a place process, said method comprising the steps of:
causing, in response to a meet instruction issued by said first agent process, execution of a procedure by said second agent process wherein said procedure is a portion of said second agent process and comprises a collection of computer instructions contained within said second agent computer process;
and providing to said first agent process, in response to a second instruction which is issued by said second agent process and which is part of said procedure, means for accessing said second agent process.
162. The method of Claim 161 wherein said step of causing execution of a procedure by said second agent process comprises:
causing, in response to said meet instruction issued by said first agent process, execution of a second procedure by said place process wherein execution of said second procedure by said place process issues an instruction which causes said second agent process to execute said first-mentioned procedure.
163. The method of Claim 161 wherein said means for accessing said second agent process is provided to said first agent process by said place process.
164. The method of Claim 163 wherein, at approximately the time said place process provides to said first agent process said means for accessing said second agent process, said place process provides to said second agent process means for accessing said first agent process.
165. A method for transferring a first computer process from a first computer system to a second computer system, said first computer system comprising a first CPU
and a first memory and said second computer system comprising a second CPU and a second memory, said method comprising:
initiating execution of said first computer process within said first CPU wherein said computer process has an execution state;
suspending execution of said first computer process within said first CPU; -representing said first computer process as data in said first memory wherein said data includes said execution state of said first computer process at the time execution of said first computer process is suspended;
transferring said data from said first memory to said second memory;
forming a second computer process on said second computer system from said data wherein said second computer process has said state of execution represented in said data; and causing execution of said second computer process, thereby effectively simulating resumption of execution of said first computer process, within said second CPU.
Description  (OCR text may contain errors)

DEMANDES OU BREVETS VOLUMINEUX

COMPREND PLUS D'UN TOME.
CECI EST LE TOME r DE
NOTE: Pour les tomes additionels, veuillez contacter le Bureau canadien des brevets THAN ONE VOLUME
. THIS 1S VOLUME ~ OF
NOTE: For additional volumes please contact the Canadian Patent Office WO 95/02219 ~ 3 ~ '~ PCTIUS94/07397 SYSTEM AND METHOD FOR DISTRIBUTED COMPUTATION BASED UPON
' MOVEMENT, EXECUTION, AND INTERACTION OF PROCESSES IN A NETWORK
A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent files or records, but otherwise reserves all copyright rights whatsoever.
FIELD OF THE INVENTION
This invention relates to a distributed computing environment and in particular, to an improved distributed computing environment wherein processes are created in an object oriented environment and direct their own movement throughout a computer network.
BACKGROUND OF THE INVENTION
Many different computing systems are available today, but most of these systems are built around fundamental components such as those illustrated in Figure 1.
Typically, a computer system 20 includes a central processing unit 10 (CPU), which is connected through an input bus 11 to an input module 12 and through an output bus 13 to an output module 14. CPU 10 is also connected _ through data buses 15, 16 to a memory unit 17.
CPU 10 provides control and computing functions.
Input and output modules 12, 14 are used to communicate between the computer user and CPU 10. Input module 12 su~;plies information to CPU 10. Typical input devices are '21~36~~
a keyboard and a mouse. Output module 14 displays information from central processing unit 10. Typical output modules include video display monitors, printers, plotters and other visual display means. Input unit 12 and output unit 14 are frequently referred to as input/output (I/O) units.
Memory unit 17 typically contains two general types of information, instructions and data. A computer program is a sequence of instructions that are executed by the computer to perform a specific function. Data in memory unit 17 are processed by CPU 10 in response to the instructions from a computer program which is executing in CPU 10. An executing computer program is called a computer process.
Memory unit 17 typically includes mass memory 17A, sometimes called secondary memory, and main memory 17B.
Main memory 17B is usually used to store at least a portion of a program currently being executed by CPU 10 and data required by the program. Mass memory 17A, e.g.
magnetic disk or tape, is used to store programs, data, or portions of either programs or data which are not needed immediately by CPU 10 or which cannot be accommodated in main memory 17B because of size limitations of main memory 17B.
The operating system of a computer is a computer process which coordinates operation of CPU 10, I/O
devices 12, 14 and 17A and main memory 17B and which provides an interface between user applications (defined below) and computer hardware. As is used herein, the term computer hardware refers to the physical components of the computer system. As an operating system is necessary for the orderly operation of a computer system, the operating system is loaded into main memory 17B and executed during power up of the computer system (a process called "booting") and remains in main memory 17B and in execution until the computer system is ultimately deactivated.

WO 95/02219 ~ PCT/L1S94/07397 A user application generally refers to a computer process which executes in addition to, and is generally considered distinct from, the operating system. A user application begins at some point in time as source code, i.e., computer instructions in a form intelligible to human beings. This source code is translated or "compiled" into object code, which is a form intelligible to computer system 20, particularly CPU 10. Several object code modules may be linked to form a single executable image, which is in a form which can be processed by CPU 10 of computer system 20. The program is executed by copying the executable image into main memory 17B and instructing CPU 10 to carry out the instructions, one by one, of the executable image. As stated above, a computer program in such state of execution is called a process.
The instructions executed by a process executing within computer system 20 generally manipulate the resources of computer system 20. For example, instructions can manipulate data stored in mass memory 17A, accept data from input module 12 or display data on output module 14. Some processes, which are designed to execute on a first computer system, are configured to issue computer instructions which manipulate resources of a second computer system.
It is well-known in the art of computer communications to connect two or more computer systems such that data can be transferred between two or more computer systems. Two or more computer systems so connected are collectively called a "network". In such a network, each computer system treats each other computer system as an I/O device to which data can be sent and from which data can be received.
A first process executing on a first computer system can issue instructions to a second process executing on a second computer system. The second process then manipulates the resources of the second computer system in accordance with the instructions received from the first process. The second process is called a "server" process because the second process provides a service to the first process, which is called a "client" process. The service provided is the manipulation of resources of the second computer system.
Many computer communication systems implement what is called "remote procedure calling". In remote procedure calling, a client process, which is executing on a first computer system and which seeks to manipulate data on a second computer system, requests action through a server process, which is executing on the second computer system and which manipulates data in response to requests from the client process. The server process performs any of an inevitably finite list of operations on behalf of the client process. Often it is the case that the precise function requested by the client process is not among the particular operations performed by the server process. To perform a function for which no server process operation is defined requires several requests to and responses from the server process that require substantial use of communications media which in turn involves considerable expense and time.
Logic flow diagram 200 (Figure 2A) illustrates an example of remote procedure calling. Logic flow diagram 200 is described in the context of computer systems 20A
and 20B (Figure 2B), which are interconnected through network 256. Suppose a client process 252 on computer system 20A is to delete all files on computer system 20B
whose names match a given pattern, e.g., file names ending in ".TMP", and which have not been modified in at least 30 days. Client process 252 sends to server process 254 which is executing on computer system 20B a request to list all files in step 21 (Figure 2A). Arrow 260 (Figure 2B) represents that request. Processing transfers WO 95/02219 ~ PCTIUS94I07397 from step 21 (Figure 2A) to step 22 in which client process 252 (Figure 2B) receives the list from server process 254, as shown as arrow 262. Steps 21 and 22 (Figure 2A) involve two transfers across network . 5 communications media.
Double rectangles in Figures 2A and 3A indicate information transfer across network 256 (Figure 2B).
Whereas execution of an instruction within either computer 20A or computer 20B takes between several nanoseconds to several microseconds, transfer of information across network 256 generally takes seconds or minutes depending on the configuration of network 256 and on the amount of information. In some networks, transfer of a large amount of data can take an hour or more.
Processing transfers from step 22 (Figure 2A) to for each filename step 23. For each filename step 23 and next steps 23a, 23b and 23c define a loop within which each filename of the list of filenames is processed by the client process. For each filename of the list, processing transfers from for each filename step 23 to a test step 24 in which client process 252 (Figure 2B) compares the filename to a pattern. If the filename does not match the pattern, processing transfers from test step 24 (Figure 2A) through next step 23a to for each filename step 23 and the next filename is processed.
Conversely, if the filename matches the pattern, processing transfers from test step 24 to step 25. In step 25, client process 252 (Figure 2B) sends to server process 254 an instruction to provide the date that the particular file was last modified as shown by arrow 264.
Processing transfers from step 25 (Figure 2A) to step 26 in which client process 252 (Figure 2B) receives from server process 254 the date requested as represented by arrow 266. Steps 25 and 26 (Figure 2A) involve two more transfers across network communications media for each such matching filename.

~1~3~~~
Processing transfers from step 26 to a second test step 27 in which client process 252 (Figure 2B) compares the date received with the current date less 30 days. If the date indicates that the file has been modified within the preceding 30 days, processing transfers from second test step 27 (Figure 2A) through next step 23b to for each filename test step 23 in which the next filename is processed. Conversely, if the date indicates that the file has not been modified within the preceding 30 days, processing transfers from second test step 27 to step 28.
In step 28, client process 252 (Figure 2B) sends to server process 254 an instruction directing server process 254 to delete the file from computer system 20B as represented by arrow 268. Processing transfers from step 28 to step 29 (Figure 2A), in which client process 252 (Figure 2B) receives from server process 254 acknowledgement of the deletion as represented by arrow 270. Thus, steps 28 and 29 (Figure 2A) involve two further transfers across network 256 (Figure 2B) for each file to be deleted.
Processing transfers from step 29 through next step 23C to for each filename step 23. If all filenames in the list have been processed, processing transfers from for each filename step 23 to terminal step 20b in which processing terminates. Thus, processing according to logic flow diagram 200 (Figure 2A) requires a multitude of information transfers across network 256. Such heavy use of network communications media is both costly and time consuming.
An alternative approach to remote procedure calling is "remote programming". In remote programming, a process, which is called a "client process" and which executes on a first computer system, sends to a server process executing on the second computer system a list of instructions. The instructions are then carried out by the server process effectuating the goal of the client ~143u7?
process. The instructions which the server process is designed to carry out must have some degree of generality;
i.e., the instructions must provide some degree of decision-making capabilities.
. 5 Logic flow diagram 300 (Figure 3A) illustrates an example of remote programming. To effectuate the example given above with respect to Figures 2A and 2B in a remote programming environment, client process 352 (Figure-3B) in step 31 (Figure 3A) builds a program whose instructions are represented by logic flow diagram 31-P. Logic flow diagram 31-P is described below.
Processing transfers from step 31 to step 32, in which the program is transferred to computer 30B
(Figure 3B) through network 356 as represented by arrow 358. Processing transfers from step 32 to step 33 (Figure 3A) in which the program is executed within computer 30B. During execution, the program is process 352A (Figure 3B) within computer 30B. The instructions of program 352A are executed according to logic flow diagram 300.
In step 31-B (Figure 3A), a list of filenames is created. Processing transfers from step 31-B to for each filename step 31-C. For each filename step 31-C and next steps 31-H, 31-I, and 31-J form a loop within which each filename in the list of filenames is processed. For each filename, processing transfers from for each filename step 31-C to test step 31-D in which the filename is compared to a pattern. If the filename does not match the pattern, processing transfers from test step 31-D through next step 31-H to for each step 31-C. Conversely, if the filename matches the pattern, processing transfers from test step 31-D to step 31-E.
In step 31-E, process 352A retrieves the date of last modification of the file having the filename. Processing transfers from step 31-E to a second test step 31-F in which the date is compared to the current date less 30 WO 95/02219 ~ PCT/US94/07397 1~3~72 days. The date of last modification of the file is retrieved from server process 354 (Figure 3B). If the date of last modification is more recent than the date of 30 days prior, processing transfers from second test step 31-F (Figure 3A) through next step 31-I to for each filename step 31-C. Conversely, if the date of last modification is prior to the date of 30 days prior, processing transfers from second test step 31-F to -step 31-G. In step 31-G, the file is deleted by issuing an appropriate instruction to server process 354 (Figure 3B). Processing transfers from step 31-G
(Figure 3A) through next step 31-J to for each filename step 31-C. If all filenames in the list have been processed, processing transfers from for each filename step 31-C to terminal step 31-K in which process 352A
terminates. As indicated by arrows 360 (Figure 3B), all interaction between process 352A and server process 354 transpired entirely within computer 30B without use of network 356.
After successful completion of the program, processing transfers from step 33 to step 34 (Figure 3A) in which server process 354 (Figure 3B) reports to client process 352 that the program completed successfully as represented by arrow 362. Note that this remote programming procedure involves only two uses of network communications media: a first to send the program or list of instructions to server process 354 as represented by arrow 358 and a second to receive from server process 354 notification of successful completion of the program as represented by arrow 362. Note also that client process 352 was able to request of server process 354 an operation not explicitly provided by, and perhaps not even anticipated by the developers of, server process 354.
However, remote programming encounters the same inefficiencies of remote procedure calling when a computer process seeks to coordinate manipulation of data on two or _g_ __ WO 95/02219 ~ ~ PCT/US94/07397 more computer systems. For example, a process on computer X may seek to delete all files on computer system Y or computer system Z which have the same name, date of last modification and size as any file on computer X. Since server process 354 described above with respect to Figures 3A and 3B is capable of directly manipulating resources on the computer system on which the server process is executing, i.e. computer system Y; the server process is unable to coordinate data management spanning more than one computer system without producing the inefficiencies discussed above in conjunction with remote process calling.
An improvement over remote programming is described by C. Daniel Wolfson, et al., "Intelligent Routers", 9th International Conference on Distributed Computing Systems at pp. 371-375 (1989). Wolfson et al. disclosed a system wherein a process could move from one computer system to another within a network. The process directed its own movement through the network by issuing an instruction whose execution caused the process to be moved. The instruction specified to which computer system the process was to be moved.
However, the solution in the form disclosed in Wolfson et al. was of limited functionality. Processes were required to be configured to comport with the specific requirements of the computer system to which the respective processes moved. Processes were not able to communicate with one another; processes instead directly stored or directly retrieved data from the mass memory of the computer system within which the respective processes were executing. One process could only interact with another process only if both processes knew the precise location within mass memory where~messages were to be stored and/or retrieved.
The system disclosed by Wolfson et al. did not provide any security. Thus, to the extent that the WO 95/02219 PCT/US94/07397 _ ~143s72 Wolfson et al. system allowed one process to interact with another process, a malicious process could interfere with the execution of a second process. In addition, the set of instructions provided to processes of the system disclosed by Wolfson et al. was very limited; only character string data and numerical data representing real numbers were provided for.
Another system similar to that disclosed by Wolfson et al. was that described by D. Tsichritzis et al., "KNOs:
KNowledge Acquisition, Dissemination, and Manipulation Objects", ACM Trans. on Office Information Systems, vol.
5, no. 1, pp. 96-112 (1987). The system described by Tsichritzis et al. added to the system described by Wolfson et al. the generality of object-oriented programming.
Object-oriented programming organizes data and instructions into "objects" in which data represent the "state" of an object and instructions are grouped into tasks or "operations" that the object can "perform".
Object-oriented programming represents a very useful conceptual framework in which problems solved by a computer may be more easily reduced to a series of computer instructions.
Objects created in an object-oriented environment are grouped into classes. Objects of a class have states of the same structure and perform the same operations. The system disclosed by Tsichritzis et al. did not provide for the mobility of class definitions; therefore, objects of a given class created within the environment described by Tsichritzis et al. could only travel to those computer systems for which the given class was defined.
Additionally, objects in the environment described by Tsichritzis et al. were represented in a form that required that the computer systems of the network, within which objects could travel, were homogeneous.

The system taught by Tsichritzis et al. further described an instruction by which a first process, i.e., a "head" process, could create copies of itself. The copies were called "limbs". Tsichritzis et al. taught that limbs could be sent to remote computer systems at the direction of the head process. However, the limbs were not active;
the computer instructions of a limb were executed only at the direction of the head process. Directing the activity of a limb process positioned in a remote computer system required that the head process issue commands to the limbs across network communications media, involving considerable time and expense in large networks.
Further inefficiencies were found in the system taught by Tsichritzis et al. when a small computer, e.g., a personal computer, was connected to a large network through a large computer, e.g., a mainframe computer. For example, to send many limbs to various computers of the network, the small computer was required to send many identical limbs between the small computer and the large computer. As limbs, which are copies of the head process, could be quite large, inefficiencies in such a system could have been quite substantial.
Tsichritzis et al. taught a system wherein a first process communicated with a second processes by giving to the second process a reference to the first process. A
reference is data which identifies an object and which grants access to the object identified. As the second process contained a reference to the first process, the second process could request that the first process execute specific computer instructions which were contained within the first process. However, as the first process provided the second process with a reference to the first process without simultaneously obtaining a reference to the second process, the second process had access to the first process and the first process did not have reciprocal access to the second process. Such a ~1~3~~~92 system permitted a "malicious" process, i.e., a process designed by a malicious programmer or inadvertently designed to harm other processes, to gain access to a second process without granting reciprocal access.
Thus, in spite of the achievements of Wolfson et al.
and Tsichristzis et al., neither discloses a system in which new classes of objects and processes can be created and can be transported to and processed by computer systems within the network which do not contain class definitions for the new classes. Neither does either Wolfson et al. or Tsichristzis et al. disclose a mechanism for implementing a complex and hierarchical security scheme. Furthermore, neither Wolfson et al. nor Tsichristzis et al. disclose computer processes which can interact with one another but which cannot gain access to other processes without granting reciprocal access.
Furthermore, neither Wolfson et al. nor Tsichristzis et al. disclose computer processes which can travel to several computer systems simultaneously by creating several autonomous clones of a process and transporting the clones to respective computer systems. The limbs taught by Tsichristzis et al. are not autonomous as the computer process controls the actions of limbs at remote computer systems. Furthermore, neither Wolfson et al. nor Tsichristzis et al. disclose efficient transportation of the clones to the respective computer systems in which transportation of redundant information across network communication media is minimized.
SUMMARY OF THE INVENTION
According to the principles of this invention, a set of interpreted, object-oriented computer instructions is used to create novel computer processes that are executed in a distributed computer system. A particular process that utilizes the set of computer instructions is activated by an engine that is executing within the distributed computer system. The engine interprets and effectuates the instructions of the particular process within the distributed computer system.
Each engine in the distributed computer system interprets the instructions that define a process uniformly.
In other words, the instructions which comprise a process and which are interpreted by a first engine do not depend on the particular configuration of the computer system within which the first engine is executing. The instructions, therefore, can be moved to and interpreted by a second engine, even if the first and second engines are executing within two separate computer systems whose operating systems and hardware are otherwise generally incompatible. The interpreted instructions are implemented by using interfaces to communication, storage, computing and other subsystems of the computer system within which the engine is executing.
These interfaces, which collectively form part of one embodiment of the present invention, are described in Appendix C, which is a part of this disclosure.
Two or more engines are interconnected to form a network. The network is a universe within which computer processes of this invention travel. In one embodiment, the network encompasses computer systems that would normally be considered clients of, and not part of, other networks. For example, one embodiment of the network of this invention encompasses the workstations which are connected by the network. The term "workstation" is used herein to describe a computer system which is used for purposes other than the transportation of information to other computer systems.
The engines of this invention are interconnected so that engines are capable of moving computer processes among themselves.
To transport a particular computer process, the computer process is suspended and the execution state of the computer process is preserved. The instructions of the computer process, the preserved execution state, and objects owned by the computer process are packaged, or "encoded", to generate a string of data that is configured so that the string of data can be transported by all standard means of communication used to form a network. In one embodiment, the string of data is transported between engines as specified by the protocol described in Appendix D, which is a part of this disclosure.
Once transported to a destination computer system of the network, the string of data is decoded to generate a computer process within the destination computer system.
The decoded computer process includes those objects encoded as described above and has the preserved execution state.
The destination computer system resumes execution of the computer process. The instructions of the computer process are executed by the destination computer system to perform complex operations, including defining, creating and manipulating data objects and interacting with other computer processes executed by the destination computer system. As computer processes are uniformly interpreted throughout the network and are encoded and decoded for transport between computer systems, the computer processes of this invention provide a new level of functionality and versatility in intercomputer communication.

In one embodiment, two classes built into the set of computer instructions of this invention are an agent class and a place class. Instructions formed using the agent class are interpreted by an engine to form an "agent process", sometimes referred to herein simply as "an agent"
An agent is all active object in that an engine initiates execution of an agent upon creation of the agent. The agent class provides instructions which enable an agent to (i) examine and modify itself, (ii) transport itself from a first place process, which is described more 14a ~, ~ ''~
completely below, in the network to a second place process in the network, and (iii) interact with other agents found at the second place process. The first and second place processes can execute within two separate computer systems of the network. Thus, an agent can travel from a first computer system to a second computer system.
An agent can contain information which is carried with the agent from the first place process to the second place process. Additionally, the extensibility, generality and functionality of the set of computer instructions discussed more completely below provide tremendous flexibility and control in determining, according to the instructions which comprise an agent, how and to what destination the agent, and the information contained therein, travels.
The power of agents is counterbalanced by "permits".
A permit limits the particular capabilities of a particular agent on particular occasions. The permit of an agent specifies which of several of the operations defined for the agent class can or cannot be performed by the agent. The permit further limits the amount of processing resources the agent can consume and the time at which the agent expires. Additionally, the permit specifies the priority of execution of the agent relative to other agents.
Instructions formed using the place class are interpreted by an engine to form a "place process", sometimes referred to herein simply as "a place". A place is also an active object, and the place class provides instructions which enable a place to examine and modify itself and to serve as a venue for agents and a context in which agents can interact. Agents each occupy a respective single place. Additionally, each place can occupy a single other place.
Places provide a degree of privacy and security for agents. For example, an agent, which is configured to WO 95102219 PCT/US94I07397 __ ~z1 ~3~~ ~z avoid contact with other agents, can occupy a place which is not generally known to other agents, or which denies ingress to other agents. Conversely, an agent, which is configured to provide services publicly to a large number of agents, can occupy a place which is widely known to other agents and which grants ingress to other agents.
Agents cannot interact at a distance; i.e., no two agents can interact unless both occupy the same place. To interact with a second agent occupying a second place, a first agent, which occupies a first place, issues a command causing the first agent to be transported to the second place as discussed below with respect to the "go"
operation. While both the first and second agents occupy the second place, the first agent can interact with the second agent.
Thus, the processes of this invention are a novel implementation of "remote programming," and not the more familiar "remote procedure calling" paradigm. Remote programming improves upon remote procedure calling by (i) enabling processes to interact without communicating across network communications media and (ii) improving the performance of the interactions between processes by eliminating communication across network communications media which often has high-latency.
The movement of an agent process, from a first place process in a first computer system to a second place process either in the first computer system or in a second computer system is called a trip. The agent initiates the trip by using a "go" operation, which is defined in the agent class. The agent controls the movement either through the network or within the computer system by creating and submitting, as an argument to the "go"
operation, a ticket which defines the trip. In one embodiment, the ticket specifies the place to which the agent is to travel, the "way" by which the agent is to travel, the amount of time in which the trip must be WO 95/02219 ~ ~ PCT/US94/07397 completed, and an indication of the urgency of the trip, i.e., the priority of the trip relative to other trips by other agents that may be concurrently scheduled.
The ticket can identify the destination place by 0 5 specifying the address, name, class or any combination of the address, name and class of a place to which the agent is to travel. The destination of the trip is any place of the specified address, name and/or class which grants the agent ingress within the time permitted by the ticket.
In the "go" operation, the agent is moved from a first place to a second place as follows: (i) the agent is suspended and the execution state of the first agent is preserved; (ii) the agent is represented in a standardized form, i.e., an octet string, which includes representation of the agent's preserved execution state; (iii) the standardized form is transported from the first place to the second place, potentially involving the transfer of the standard form from a first computer to a second computer; (iv) the agent at the second place is formed from the standardized form, including the execution state represented in the standardized form; and (v) interpretation of the agent is resumed, the agent initially having the execution state represented in the standardized form.
As mentioned above, the set of computer instructions of the present invention is object=oriented. Therefore, ail objects formed according to the present invention are organized into classes. All classes in the present invention are represented by data objects which can be moved along with an agent. Thus, classes, which are not defined within a place, can nevertheless be used by an agent travelling to the place simply by the agent defining the classes and transporting the corresponding class objects to the place.
Every object in one embodiment of the present invention is owned by a process, i.e., either an agent or wo 9s~0~~14 3 6 '~ 2 a place. When an agent travels from a first place to a second place, all objects owned by the agent are effectively moved to the second place along with the agent. However, as discussed below, objects which are equivalent to objects already occupying the second place are not transported in one embodiment of the present invention. In addition to the objects owned by the agent, all class objects defining classes of which those objects are members are moved along with the agent to the second place. A class object is an object constructed in accordance with the set of computer instructions described below and in Appendix A which represents a class of objects.
Thus, if an object owned by an agent travelling to a second place is a member of a class not defined in the second place, the object can still travel with the agent as the agent also carries to the second place class objects defining the classes of which the object is a member. In the prior art, a process which travelled from a first computer to a second computer could only process data objects which belonged to classes defined on the second computer. Class definitions were not typically transported with migrating processes in the prior art. As new classes can be formed within a first place and members of those new classes are free to travel to other places for which those new classes are not defined, the present invention provides agents a level of extensibility, mobility and generality not found in the prior art.
In the present invention, many classes of objects are defined at multiple places and therefore do not need to be moved to such places. Class objects and any other objects, which are likely to be found at many places are made interchangeable in one embodiment of the present invention.
Interchangeable objects each have a digest. An interchangeable object, which is owned by an agent that is _lg_ 2~~3s~~
travelling from a first place to a second place, does not travel with the agent. Rather, the digest of the interchangeable object is moved with the agent to the second place. When the agent arrives at the second place, interchangeable objects in the computer system which contains the second place are examined to determine whether any of the interchangeable objects has a digest equal to the digest transported with the agent. If-such an interchangeable object is found in the computer system which contains the second place, the interchangeable object is substituted for the interchangeable object, which was left behind at the first place.
If, however, no such interchangeable object is found at the second place, the interchangeable object left behind at the first place is moved to the second place.
In this way, the movement of objects across network communications media is avoided when equivalent objects are present at the destination place. In particular, as class objects are interchangeable, an agent travelling to a second place causes the movement to the second place of only those class objects defining classes which are not defined within the second place. Avoiding unnecessary movement of class objects through the network makes practical the movement of objects owned by an agent to places which contain no definition of one or more classes to which those objects belong.
In one embodiment of the present invention, classes are identified by citations. In a computer network having many computer systems supplied by different vendors, various versions of the disclosed instruction set can be implemented on various computer systems to which and from which agent processes can travel. A citation is an object which is used to identify classes or objects which are forward-or backward-compatible with one another. Therefore, an instruction requiring use of an object of a first class, or the first class itself, can use an object of a second ~143~'~2 class, or the second class itself, depending on the particular requirements of the instruction, if the second class is backward-compatible with the first class.
An agent, occupying a first place, is also capable of creating one or more clones of the agent and moving each clone to a respective place. A clone is an agent process which is the result of duplicating an existing agent process. An agent initiates and controls the creation of one or more clones, and the movement of each clone to a respective place process by performance of operation "send." The agent specifies the number of clones to be created and defines the corresponding trips by creating one or more tickets which are supplied as arguments to operation "send". Each ticket defines a trip to be taken by a respective clone of the agent. The number of tickets supplied by the agent defines the number of clones created.
Once a clone of the agent is moved to a second place, the clone initially has the execution state of the agent at the time the clone was created. Thus, when the clone arrives at a second place, the clone continues to execute so as to simulate the movement of the agent to the second place as described above with respect to operation "go."
In the prior art, a first process, i.e., a head process, created limb processes which had the same interface, i.e., could generally perform the same operations, as the head process. Limb processes did not act except as directed by the head process and could move to remote computer systems only at the direction of the head process.
In the present invention, however, a clone is autonomous and is not controlled by the agent which created the clone. For example, the clone can be configured to ignore all attempts by the agent to arrange a meeting, which is discussed below, whereby the agent and the clone can interact. The agent must attempt to interact with a clone of the agent in the same way the agent attempts to interact with any other agent, as discussed below. As the clone is autonomous, the clone occupying the second place can travel to a third place by r 5 performance of operation "go" without being so instructed by the agent and even without the consent of the agent.
As limb processes of the prior art were not active and acted only at the direction of a head process, the head process and a remote limb process necessarily interacted across network communications media. The paradigm of this prior art system is therefore more closely related to remote procedure calling. In contrast, clones of agents formed in accordance with the present invention are active and autonomous and embody all of the instructions which form the agent. Therefore, no interaction is required (in fact, no interaction is allowed) across network communications media. Therefore, the paradigm of the present invention is more closely related to remote programming. The present invention therefore represents a significant increase in generality over the prior art.
Increases in efficiency in one embodiment of the present invention are realized in performance of operation "send" by deferring cloning of an agent as long as possible. For example, a first clone and a second clone are to be sent to a first place and a second place, respectively. Suppose that the agent is executing within a first computer system, that the first place is executing within a second computer system, and that the second place is executing within a third computer system. Suppose further that in travelling to the first and second places, the first and second clones must travel through an engine executing on a fourth computer system. In such a case, a single clone is formed and transported to the engine executing within the fourth computer system. Thus, only a single clone is created within the engine interpreting the original agent thereby saving space within that engine and only a single clone is transported across network communications media to the fourth computer system thereby saving time in transporting clones to the fourth computer system. The engine executing within the fourth computer system forms from the single clone the first and second clones and transports the first and second clones to the first and second places, respectively.
As an agent can own objects which account for a substantial majority of the size of the agent, e.g., a rasterized graphical image such as a facsimile transmission or digitized sound, avoiding sending several copies of such large objects, each copy owned by a respective clone, to a single engine saves substantial time and expense.
A first agent, occupying a place, can initiate a meeting between the first agent and a second agent occupying the place. During such a meeting, the first agent can transfer to and receive from the second agent data in the form of objects, and the second agent process can transfer to and receive from the first agent data in the form of objects.
The present invention represents a significant improvement over a prior art system which teaches the posting of messages on a virtual bulletin board by a first process. The messages are then "read" by the intended recipient process. In the prior art system, a first process gives to a second process a reference to the first process by posting the reference on the virtual bulletin board. However, since there is no mechanism for simultaneous exchange of references, the second process has access to the first process before giving to the first process a reference to the second process. Thus, there is no mechanism to prevent "malicious" processes from gaining access to other processes without granting to the other processes access to themselves.

__ WO 95/02219 ~ ~ ~ J ~ ~ ~ PCTIUS94/07397 The present invention represents an improved method of establishing contact between two processes such that a first process cannot obtain access to a second process without simultaneously granting to the second process access to the first process.
In one embodiment of the present invention, two agents can interact only if both agents occupy the same place and the place is a meeting place. A meeting place is a place that is a member of a class of meeting places, which is a subclass of the class of places. A first agent directs that a meeting be arranged between the first agent and a second agent by issuing an instruction directing the meeting place to arrange the meeting. The issued instruction is called operation "meet", and the first agent's issuance of the instruction is called "requesting a meeting".
The first agent supplies, as an argument to the instruction, a petition defining the meeting. The petition defines the meeting by specifying the second agent as the petitioned agent. The second agent is specified by specifying the name and/or the class of the second agent.' The petition further defines the meeting by specifying the amount of time in which the meeting must be arranged or abandoned.
In arranging the meeting, the meeting place supplies to the second agent the name and class of the first agent and indicates that the first agent has issued an instruction requesting a meeting with the second agent.
The second agent examines the name and class of the first agent and responds to the meeting place either accepting or rejecting the meeting with the first agent.
If the meeting is rejected, the first agent is informed that the second agent is not available. If the meeting is accepted, the meeting place gives to the second agent a reference to the first agent and gives to the first agent a reference to the second agent. With a reference to the second agent, the first agent (i) can direct the second agent to perform operations; (ii) can supply to the second agent objects as arguments; and (iii) can receive from the second agent objects as results.
As the second agent has a reference to the first agent, the second agent has similar capabilities with respect to the first agent.
Either the first or the second agent can terminate the meeting between the two by issuing an appropriate command. The meeting place, in performing an operation "part" in response to the issued command, voids any references to the second agent contained within the first agent and voids any references to the first agent contained within the second agent, thereby terminating interaction between the two agents.
The present invention represents a significant improvement over the prior art as a first agent cannot gain access to a second agent unless the second agent agrees or without granting to the second agent access to the first agent. Additionally, since two agents cannot interact unless both occupy the same meeting place, intermediate levels of security are available to agents. For example, an agent can protect itself from other agents by occupying a meeting place whose location is not widely known by other agents. Inversely, an agent, which is designed to be highly visible, may occupy a well-known meeting place. No such mechanism is available in the prior art.
According to one aspect the invention provides a method for implementing remote programming in a computer network comprising the steps of: defining a plurality of object-oriented classes including an agent class and a place class; forming instructions for a computer process, said instructions including said object-oriented classes, subclasses of said object-oriented classes, and a go operation; and interpreting said instructions on a processor in said computer network wherein, in response to said go operation, an agent process is transported to a place process and further wherein said agent process is a member of said agent class and said place process is a member of said place class.
According to another aspect the invention provides a method for implementing remote programming in a computer network comprising the steps of: defining a plurality of object-oriented classes including an agent class and a place class; forming instructions for a computer process including said object-oriented classes, subclasses of said object-oriented classes, and a send operation; and interpreting said instructions on a processor in said computer network wherein, in response to said send operation, a clone of an agent process is transported to a place process and further wherein said clone and said agent process are each a member of said agent class and said place process is a member of said place class.
According to another aspect the invention provides a method for interprocess communication in a computer comprising the steps of: defining a plurality of object oriented classes including an agent class; forming instructions for a computer process including said object-oriented classes, subclasses of said object-oriented classes, and a meet operation; and interpreting said instructions on a processor in said computer network wherein, in response to said meet operation, a meeting place process provides a first agent process access to a second 24a agent process and provides said second agent process access to said first agent process and further wherein said first and second agent processes are members of said agent class.
According to another aspect the invention provides a method for controlling movement of processes in a computer network, said method comprising the steps of: defining a plurality of object-oriented classes including an agent class, a place class, and a ticket class; forming a plurality of place processes in said computer network wherein each of said plurality of place process is a member of said place class; forming an agent process wherein said agent process is a member of said agent class and occupies a first place process in said plurality of place processes;
and forming a ticket wherein said ticket is a member of said ticket class and defines a trip involving the movement of said agent process from said first place process to a second place process in said plurality of place processes.
According to another aspect the invention provides a method for limiting capabilities of processes in a computer network, said method comprising: defining a plurality of object-oriented classes including a process class and a permit class; forming a process wherein said process is a member of said process class; and forming a permit wherein said permit is a member of said permit class and specifies one or more capabilities of said process.
According to another aspect the invention provides in a computer, a method of interpreting processes of various versions of an instruction set, said method comprising:
defining a plurality of object-oriented classes including a class of classes and a class of citations; forming one or more class objects wherein said class objects are members of 24b said class of classes; forming within a first of said class objects a citation wherein said citation specifies said first class object and specifies which of said class objects are backward compatible with said first class object.
According to another aspect the invention provides in a computer network having a plurality of computers, a communication process comprising: providing a plurality of place processes within said computer network wherein each place process is a locale in one of said computers for zero or more agent processes; specifying, by a ticket means, a trip for an agent process to a destination place process in said plurality of place processes; and transporting, in response to a send operation within said agent process, a clone of said agent process to said destination place process.
According to another aspect the invention provides in a computer, a communication process comprising: providing a first agent process and a second agent process; specifying a meeting between said first and second agent processes by a petition means; and arranging said meeting between said first and second agent processes as defined by said petition means.
According to another aspect the invention provides in a computer network having a plurality of computers, a communication system comprising: an agent means having a ticket means and a send operation; and a plurality of place means wherein each place means is operative in one of said plurality of computers; wherein said agent means is at a first place means in said plurality of place means; said ticket means specifies a trip for said agent means to a destination place means in said plurality of place means;
24c and said send operation transports a clone of said agent means to said destination place means.
According to another aspect the invention provides in a computer, a communication system comprising: meeting place means having a meet operation wherein said meeting place is a locale for a plurality of agent means; a petition means; a first agent means in said plurality of agent means;
and a second agent means in said plurality agent means wherein; said first and second agent means can meet with any of said plurality of agent means; said petition means specifies a meeting between said first and second agent means; and in response to said meet operation, said meeting place means arranges said meeting.
According to another aspect the invention provides in a computer network having one or more computers, a method for transferring data from a first engine process to a second engine process, said method comprising the steps of:
(a) providing means for executing one or more agent processes wherein said executing means comprises said first and second engine processes and further wherein each said agent process comprises instructions from a computer instruction set and has an execution state; (b) providing within said computer instruction set a go instruction wherein said go instruction is contained within a first agent process executing within said first engine process and further wherein performance of said go instruction causes:
(i) suspension of execution of said first agent process by said first engine process; (ii) representation of said first agent process such that said execution state of said first agent process is preserved; (iii) transfer of said representation of said first agent process from said first engine process to said second engine process; and 24d (iv) resumption of execution of said first agent process by said second engine process; (c) causing execution of said agent process by said first engine process, thereby causing performance of said go instruction.
According to another aspect the invention provides a method for transferring data from a first engine process to one or more engine processes, said method comprising the steps of: (a) providing means for executing a plurality of agent processes which comprise instructions from a computer instruction set wherein said agent process execution means comprises said first and one or more engine processes and further wherein each said agent process has an execution state; (b) providing within said computer instruction set a send instruction wherein said send instruction is contained within a first agent process in said plurality of agent processes and further wherein, execution of said send instruction, includes the steps of: (i) forming one or more copies of said first agent process wherein said copies preserve and include said execution state of said first agent process; (ii) transferring each of said copies of said first agent process from said first engine process to a respective one of said one or more engine processes; and (iii) effectuating execution of each of said copies of said first agent process by said respective one of said one or more engine processes so as to simulate the resumed execution of said first agent process.
According to another aspect the invention provides a method of transferring data from a first agent process which is executing within a computer system to a second agent process which is executing within said computer system wherein said first and second agent processes are occupants of a place process, said method comprising the steps of:
24e causing, in response to a meet instruction issued by said first agent process, execution of a procedure by said second agent process wherein said procedure is a portion of said second agent process and comprises a collection of computer instructions contained within said second agent computer process; and providing to said first agent process, in response to a second instruction which is issued by said second agent process and which is part of said procedure, means for accessing said second agent process.
According to another aspect the invention provides a method for transferring a first computer process from a first computer system to a second computer system, said first computer system comprising a first CPU and a first memory and said second computer system comprising a second CPU and a second memory, said method comprising: initiating execution of said first computer process within said first CPU wherein said computer process has an execution state;
suspending execution of said first computer process within said first CPU; representing said first computer process as data in said first memory wherein said data includes said execution state of said first computer process at the time execution of said first computer process is suspended;
transferring said data from said first memory to said second memory; forming a second computer process on said second computer system from said data wherein said second computer process has said state of execution represented in said data; and causing execution of said second computer process, thereby effectively simulating resumption of execution of said first computer process, within said second CPU.
24f BRIEF DESCRIPTION OF THE DRAWINGS
Figure 1 shows the structure of computer systems of the prior art.
Figure 2A is a logic flow diagram of a prior art method using remote procedure calling.
Figure 2B shows a prior art network implementing remote procedure calling.
24g _ ~143~'~2 Figure 3A is a logic flow diagram of a prior art method using remote programming.
Figure 3B shows a prior art network implementing remote programming.
Figures 4A and 4B show the movement of an agent process through a network constructed in accordance with the present invention.
Figures 5A and 5B show processes, formed in -accordance with the principles of the present invention, executing within a computer system.
Figure 5C shows a portion of a class hierarchy graph in accordance with one embodiment of the present invention.
Figures 6A-6C show the movement of an agent process through a network formed in accordance with the present invention.
Figures 7A-7E show the movement of clones of an agent process through a network formed in accordance with an aspect of the present invention.
Figures 8A-8F illustrate interaction between two agent processes by using operation "meet".
Figures 9A-9E illustrate the interrelations of place processes formed in accordance with one aspect of the present invention.
Figure 10 shows the structure of a computer network formed in accordance with the principles of the present invention.
Figures 11A and 11B show alternative representations of the network shown in Figure 10.
Figure 12 is a diagram illustrating the class relationships of an agent process in one embodiment of the present invention.
Figures 13A and 13B illustrate the state of an agent process formed in accordance with the present invention immediately before and after, respectively, movement of WO 95/02219 PCTIUS94107397 _..
~1~367~
the agent process within a computer network of the present invention by performance of operation "go".
Figures 14A-14G are logic flow diagrams which illustrate the steps taken to move an agent process through a network in accordance with the present invention by performance of operation "go".
Figures 15A-15E show the structure of a network formed in accordance with the present invention and-illustrate the movement of an agent process through the network.
Figure 16 shows the structure of an agent process including a permit.
Figure 17 shows the structure of the permit of Figure 16.
Figure 18A shows the structure of the ticket of Figure 13A including a teleaddress, a citation, a telename, and a way.
Figure 18B shows the structure of the way of Figure 18A.
Figure 19 shows the structure of the telename of Figure 18A.
Figure 20A shows a dictionary formed according to one embodiment of the present invention.
Figure 20B shows the structure of a finder constructed according to one embodiment of the present invention.
Figure 21 shows the structure of an encoded agent in accordance with one embodiment of the present invention.
Figures 22A and 22B show the state of an agent process immediately prior to and following, respectively, performance of operation "entering".
Figure 22C is a logic flow diagram illustrating the steps taken in performance of operation "entering" in accordance with one embodiment of the present invention.

Figures 23A and 23B show a portion of the state of a process immediately preceding and following, respectively, performance of operation "exiting".
Figures 24A-24C are logic flow diagrams illustrating the steps taken in transporting an agent process through a network in accordance with an aspect of the present invention.
Figure 24D shows the structure of a repository of interchanged objects which is used in the steps shown in Figures 24A-24C.
Figure 25 shows a computer network formed in accordance with the principles of the present invention.
Figures 26A and 26B show a portion of the state of an agent process immediately prior to and following, respectively, performance of operation "send".
Figure 26C shows the state of a clone of the agent process immediately following performance of operation "send".
Figures 27A and 27B show the state of a engine process before and after, respectively, formation of clones of an agent process in performance of operation "send".
Figure 28 shows the computer network of Figure 25 in which clones of the agent process have travelled to respective computer systems of the computer network in performance of operation "send".
Figures 29A-29E show a computer network through which clones of an agent process travel according to another aspect of the present invention referred to as "deferred cloning".
Figure 30A shows the structure of a send frame including a nil and a list of tickets.
Figure 30B shows the structure of an encoded clone of an agent process formed in accordance with an aspect of the present invention.

WO 95/02219 ~~ ~ PCT/US94/07397 _ ~.~ 143 Figures 31A and 31B show the state of a process immediately preceding and following, respectively, performance of operation "meet".
Figure 32 is a logic flow diagram of the steps taken in performance of operation "meet".
Figure 33 shows a portion of the state of an agent process including a set of contacts.
Figures 34A and 34B show the state of a process immediately preceding and following, respectively, performance of operation "meeting".
Figure 35 is a logic flow diagram illustrating the steps taken in performance of operation "meeting" in accordance with one embodiment of the present invention.
Figure 36 shows the state of two agent processes which are interacting in accordance with an aspect of the present invention.
Figures 37A and 37B show a portion of the state of an agent process immediately preceding and following, respectively, performance of operation "part".
Figure 38 shows the state of the agent processes shown in Figure 36 immediately following performance of operation "part".
Figures 39A-39F show a portion of the state of a first agent process during interaction with a second agent process in accordance with the present invention.
Figure 40 is a logic flow diagram illustrating steps taken by the second agent process during the interaction shown in Figures 39A-39F according to one embodiment of the present invention.
Figure 41A shows the structure of a class definition in accordance with one aspect of the present invention.
Figure 41B shows the structure of a class object formed in accordance with one embodiment of the present invention.

~1~~67~

Figure 41C shows the structure of a class object formed in accordance with a second embodiment of the present invention.
Figure 42 shows the structure of an interface object formed in accordance with the present invention.
Figure 43 shows the structure of a feature definition including a set, an identifier, and a boolean.
Figure 44 shows the structure of an attribute definition including a constraint and a boolean.
Figure 45 shows the structure of an operation definition including a constraint and a list, which in turn includes a constraint.
Figure 46 shows the structure of a constraint.
Figure 47 shows the structure of an implementation object which includes two lists and six lexicons.
Figure 48 shows the structure of a method which includes a procedure and a list, which in turn includes an ident i f ier .
Figure 49 shows the structure of the procedure of Figure 48.
Figure 50 shows the structure of a citation which includes a telename, two integers, and an identifier.
Figure 51 shows the structures of two cited objects, each of which includes a respective citation.
Figure 52A is a logic flow diagram illustrating the steps taken in performance of operation "do".
Figure 52B shows the structure of a predefined frame which includes an integer and a procedure.
Figure 53 shows the execution state of a process which includes a stack, which in turn includes a user frame .
Figure 54 shows the state of the user frame of Figure 53.
Figures 55A and 55B combine as shown in Figure 55C to form a single logic flow diagram illustrating the steps taken in performance of a user-defined operation.

~~43672 Figure 56A shows the execution state of a process which includes a stack, which in turn includes a frame.
Figure 56B shows the execution state of the process of Figure 56A and a second frame.
Figure 56C shows the execution state of the process of Figures 56A and 56B, including the stack, which in turn includes the first-mentioned frame and the second frame.
Figure 57 is a logic flow diagram illustrating-the steps taken in selecting a feature definition and a method from the class hierarchy in accordance with the present invention.
Figure 58A is a diagram showing the hierarchy of the classes of which a list is a member.
Figure 58B is a diagram showing the hierarchy of the classes of which class "List" is a member.
Figure 59 is a logic flow diagram illustrating the steps taken in the initialization of an object.
Figure 60 is a logic flow diagram illustrating the steps taken in performance of operation "if".
Figure 61 is a logic flow diagram illustrating the steps taken in performance of operation "either".
Figures 62A and 62B show a portion of the execution state of a process immediately preceding and following, respectively, performance of operation "select".
Figure 63 is a logic flow diagram illustrating the steps taken in performance of operation "select".
Figure 64 shows the state of a predefined frame which represents the dynamic state of a performance of operation "select".
Figure 65 is a logic flow diagram illustrating the steps taken in performance of operation "while".
Figures 66A and 66B show a portion of the execution state of a process immediately prior to and immediately following, respectively, performance of operation "catch".
Figure 67 is a logic flow diagram illustrating the steps taken in performance of operation "catch".

WO 95/02219 ~ '~ PCTIUS94/07397 Figure 68 is a logic flow diagram illustrating the steps taken in performance of operation "loop".
Figure 69 shows the structure of a repeat frame which includes an executed object and two integers.
Figure 70 is a logic flow diagram illustrating the steps taken in performance of operation "repeat".
Figure 71 shows the execution state of a process which includes a stack which in turn contains, from-top to bottom, a user-defined frame, a predefined frame, a repeat frame, and a second user-defined frame.
GLOSSARY OF TERMS
"Abstract Class": A class which is abstract has no instances. An abstract class can have subclasses, and an abstract class can define features, methods, and properties which are inherited by the subclasses of the abstract class.
"Agent": An agent is a process which occupies a place and which is mobile, i.e. can move from a first place to a second place.
"Arguments": An argument is an object "consumed" by performance of an operation as input data. In other words, an argument is an object transferred from the requester to the responder immediately prior to performance of the operation by the responder at the request of the requester.
"Attribute": An attribute is a feature which either retrieves or sets information regarding the internal state of an object. Usually the information pertains to the object itself, but sometimes the information pertains to the reference by which the object is identified in invoking the attribute. An attribute is a pair of ~~~36~~
operations in which one sets, and the other retrieves, information regarding the internal state of the responder.
"Authority": An authority is an entity which owns and controls various resources in the network. An example of an authority is a user of the network. Authorities are created administratively and cannot be created programmatically, i.e., cannot be created at the request of a process.
"Class": A class defines (i) zero or more properties, which define the internal states of the members of the class, (ii) zero or more methods, which define the internal behavior of the members of the class, and (iii) zero or more features, which define the external behavior of the members of the class.
"Concrete Class": A class which is concrete can have instances.
"Engine Place": Every engine contains exactly one engine place which represents the engine itself.
"Engine": An engine is a machine in a computer system which manages objects, primarily processes, and executes instructions. An engine is typically a computer process executing within a computer system in addition to an operating system and various user applications. One or more engines can be executing within each computer system of a network. Each engine processes at least one place.
"Exception": An exception is an object "thrown" by performance of a feature if the feature fails to be performed completely and successfully. An exception, as the term is used herein, is alternatively a condition which causes such an object to be thrown. The responder is said to "throw" an exception, rather than "produce" an exception, because an exception arises from the failure of an operation and is therefore distinct from a result which is produced by a successful operation. The distinction between producing a result and throwing an exception is described in more detail below and in Appendix A.
"Feature": A feature is a task that an object can be directed to perform. The task is carried out by a method, which includes a set of computer instructions. Performance of a feature is accomplished by execution of the computer instructions of the feature's method. A feature is associated with a specific class of objects, and performance of a feature can vary with the specific internal state of the object performing the feature. Features are conceptually divided into the two categories: (i) attributes and (ii) operations.
"Frame": A frame is an object which records the dynamic state of a method implementing a feature during performance of the feature. A frame is used by an engine to maintain information regarding a method which the engine is executing, including information identifying the object performing the feature implemented by the method and the particular instruction that is currently executing.
"Identifier": An identifier is an object which can reference a second object. The "text" of an identifier is a string which distinguishes the identifier from other identifiers within a particular scope. The various scopes of identifiers are discussed in greater detail in Appendix A.

WO 95102219 PCT/US94/07397 __ ~143~~2 "Implementation": An implementation is a set of computer steps performed in performance of a particular feature.
An "implementation object" is an object defining the various implementations of the various features of a class.
"Instance": An object is an instance of a class if the object is a member of that class and is not a member of any subclass of that class.
"Interface": An interface defines a particular attribute or the particular arguments consumed and the result produced by performance of a particular operation. An "interface object" is an object defining the various interfaces of the various features of a class.
"Member": An object is a member of the class of which the object is an instance and any superclasses of that class.
"Method": A method is a set of computer instructions whose execution constitutes performance of a particular feature. A "method object" is an object defining a method. A method has a dynamic state during performance of the method, i.e., execution of the instructions of the method. The dynamic state of a method is represented by a frame .
"Network": All engine places, the computer systems in which the engine places execute and communications apparatus connecting those computer systems collectively form a network.
"Object": An object is an element in a computing environment within a computer system. An object has an internal state defined by zero or more properties, an internal behavior defined by zero or more methods, and an external behavior defined by zero or more features.
"Operation": An operation is a feature for which an interface and implementation is defined.
"Place": A place is a process which is a locale for zero or more processes. A place can occupy a second place.
The first-mentioned place is a subplace of the second place, and the second place is a superplace of the first place.
"Primitive": A primitive is an object that can be used in the formation of a procedure or a method, and that therefore can serve as an instruction.
"Process": A process is an object which constitutes an autonomous computation. A process is autonomous because a process performs a method without being requested to do so by another object. A process commences performance of a central method upon creation of the process, and the process is destroyed upon completion of the central method. Every object is owned by exactly one process.
Every process is owned by itself. As used herein, the term "computer process" refers to a series of instructions carried out by a computer system, which is a more general and well-known definition.
"Property": A property is an object which represents a part of the internal state of a second object.
"Reference": A reference is a data structure which identifies a particular object. An object, in being directed to perform a feature, is identified by a reference to the object supplied by the requester of the feature. References are either protected or unprotected.

WO 95/02219 PCT/US94/07397 _.
An object cannot be altered or modified using a protected reference to identify the object.
"Region": A region is one or more engine places within a network which are controlled by a single authority. A
region is generally distinguished by the close coordination and management of computer systems which support the engine places of the region. The transportation of agents within a region is therefore generally quicker and less expensive than transportation of agents between regions. A region can be, for example, a local area network which is connected to a wide area network.
"Requester": A requester is an object which directs another object, i.e. the responder, to perform a feature.
The requester supplies zero or more objects as input data to the responder of the feature requested and receives zero or one object as output data from the responder of the feature. Directing an object to perform a feature is alternatively called "requesting" the object to perform the feature.
"Responder": A responder is an object performing a feature at the direction of another object, i.e. the requester. The responder receives zero or more objects as input data and supplies zero or one object as output data in performing the feature. A responder is alternatively called a "responding" object.
"Result": A result is an object "produced" by performance of an operation as output data. In other words, a result is an object transferred from the responder to the requester immediately following performance of an operation by the responder at the request of the requester.

~143~72 "Subclass": A subclass inherits methods, properties, and/or feature definitions from one or more classes. A
subclass can define one or more properties, methods, and/or features which are not inherited from any other class and can reimplement a feature which is inherited from another class by defining a new implementation of the feature.
"Superclass": The classes from which a subclass inherits properties, methods, and/or features are superclasses of the subclass.
"Virtual Place": Every place which is not an engine place is a virtual place.
DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT
According to the principles of this invention, a novel set of computer processes are used to route a particular computer process to a selected computer system contained in a plurality of interconnected computer systems and to execute the particular computer process on the selected computer system. The computer processes of this invention are defined, in one embodiment, in terms of object-oriented computer processes. The computer processes of this invention include (i) an instruction set that defines the various operations in the process and (ii) an engine that interprets the instruction set and controls operation of a central processing unit (CPU) in performance of the computer processes.
As explained more completely below, the particular computer process directs its own movement, i.e., transportation, through the computer network by executing an instruction which identifies a destination computer system within the network and directs that the particular computer process be transported there. While executing within the destination computer system, the particular ~1436'~2 computer process may have access to information which is not available elsewhere in the network. The particular computer process can access that information and use that information to determine to which other computer systems to travel. As explained more completely below, the present invention provides computer processes with a level of mobility, extensibility and generality not found in the prior art.
In remote programming paradigms typically found in the prior art, a computer process does not direct its own movement, but is sent from a source computer system to execute within a destination computer system. If the computer process cannot direct its own movement throughout the network, the computer process cannot select further destination computer systems to which to travel on the basis of information obtained by the computer process at the destination computer system. Thus, such a computer process must return to the source computer system with the information obtained before being sent to a second destination computer system. As computer processes of the present invention can direct their own movement through the network, computer processes of the present invention represent a substantial extension of the traditional remote programming paradigm.
Existing remote programming systems, which provide processes which direct their own movement through a computer network, lack generality and extensibility. Such systems lack generality in that such systems (i) are limited to homogeneous computer networks or (ii? require that a travelling process in a heterogeneous network contain instructions which conform specifically to whichever computer system is executing the process, i.e.
computer systems to which the travelling process travels.
In the present invention, a process travels within a heterogeneous computer network and the execution of the instructions of the process is independent of which - WO 95102219 ~ 1 ~ 3 ~ '~ ~ PCT/US94/07397 particular computer system within which the process executes, i.e., to which the process travels. As discussed below in greater detail, the computer instructions disclosed below are implemented uniformly by 5 each of the computer systems of a heterogeneous network.
Thus, a process formed in accordance with the present invention can travel to any computer system within a heterogeneous computer network and execute there without any specific information regarding the nature of the computer system.
Remote programming systems of the prior art typically lack extensibility in that a process cannot create a class of data objects and utilize data objects of that class within computer systems to which the process subsequently travels. Many computer processes formed today are "object-oriented". As defined above in the Glossary of Terms, an object has (i) an internal state defined by a number of properties, (ii) an internal behavior defined by a number of methods, and (iii) an external behavior defined by a number of features. The properties, methods, and features of a group of objects with like properties, methods, and features are defined by a class. In prior art systems, a process which defines a class cannot use objects of the class while executing in computer systems to which the process subsequently travels. Transportation of class definitions with travelling processes is not feasible or practical in prior art systems because class definitions are not standardized; class definitions are generally very large and complex; and a single process generally uses objects of many classes.
Class definitions are not "standardized" in that prior art systems are typically based on existing programming languages which are not designed with mobile processes in mind. Therefore, class definitions in a process formed using one of these existing programming languages typically rely on information specific to the WO 95102219 PCTIUS94107397 _.
~1~36'~2 computer system within which the process is created, such as the specific memory addresses of portions of the class definitions. Reliance on such information, which is specific to a particular computer system, makes class definitions of prior art systems very difficult to transport to remote computer systems.
In the present invention, classes are represented by class objects which are part of the disclosed computer instruction set. A process formed in accordance with the present invention includes class objects when travelling from one computer system to another. Thus, a process can define a class of object in a first computer system, travel to a second computer system, and utilize objects of the class within the second computer system. Unnecessary delay and expense in transporting all classes utilized by a travelling process are eliminated by transporting only those classes which are not defined within the computer system to which the process travels.
Additionally, the present invention provides efficiencies not found in the prior art in transporting several clones of a process to several destination computer systems simultaneously. As discussed more completely below, substantial efficiencies are realized by deferring creation of one or more clones of the process so long as the travel path of one clone coincides with the travel path of another clone. Clones, and this aspect of the present invention, are discussed further below.
In remote programming systems, two processes share information by a first process giving a second process access to the first process. The first process gives the second process access to the first process without simultaneously obtaining access to the second process. In such a situation, the second process is free to interfere with the execution of the first process without compromising the security of the second process. In the present invention, a first process obtains access to a second process and simultaneously gives to the second process access to the first process through the cooperation and coordination of a third process. Further security is provided as the second process either accepts or rejects such an exchange with the first process according to the configuration of the second process. The third process ensures that (i) the second process agrees to such an exchange and that (ii) each process gains access to the other process simultaneously such that neither process can interfere with the execution of the other process without affording a similar opportunity to the other process.
A novel computer process of this invention is an agent process. An agent process is a computer process formed of the computer instructions disclosed below in Appendix A, which is a part of this disclosure. Agent processes are each interpreted by another novel computer process of this invention, i.e., an engine, and thereby effectuates the computer instructions disclosed below and in Appendix A. The term "interpreted" is used herein as it is understood in the art; a computer instruction in a series of computer instructions is read and executed by an engine before the next computer instruction in the series is read.
Interpreting, rather than compiling, instructions of the disclosed instruction set provides greater generality. A first agent can travel from a first computer system to a second computer system and meet with a second agent there which gives to the first agent a procedure. The first agent can then perform the procedure which the first agent was not originally designed to perform.

Yet another novel computer process of this invention is a place process. Place processes are dispersed throughout a computer network. Each agent process "occupies" a respective place process; i.e., a place 41a ~14~~b'~'~
process is part of the internal state of an agent process and therefore provides a context within which the agent process executes. Each agent process can initiate and control the agent process's transportation from a first place process to a second place process. "Agent" and "place" as used herein are shorthand terms for "agent process" and "place process", respectively.
Figure 4A shows computer network 100 which includes computer systems 120A and 1208 which are connected by communications link 102AB. Computer system 120A is executing place 220A and agent 150A. Agent 150A occupies place 220A as shown in Figure 4A. Computer system 1208 is executing place 2208. Hereinafter, the statement that an agent or place occupies a particular place should be interpreted as including a statement that the agent or place is executing within the computer system which contains the particular place.
Agent 150A issues an instruction to computer system 120A. In response to the instruction, agent 150A is transported to a place, e.g. place 2208, specified in the instruction. The instruction is called operation "go", and the issuance of the instruction by agent 150A is called herein performance of operation "go" by agent 150A.
Upon performance of operation "go" by agent 150A, (i) execution of agent 150A is suspended; (ii) agent 150A is encoded into a standardized form which preserves the execution state of agent 150A; (iii) the standardized form of agent 150A is transported to computer system 1208; (iv) agent 150A, including the preserved execution state, is decoded from the standardized form; and (v) execution of agent 150A is resumed within computer system 1208. After performance of operation "go" by agent 150A, agent 150A no longer occupies place 220A and is no longer executing within computer system 120A; instead, agent 150A occupies place 2208 and is executing within computer system 120B
(Figure 4B). By enabling an agent to travel to a remote computer system in the midst of the agent's execution, the agent is free to travel to data which the agent is configured to access.
A standardized form is a form to which agents are encoded for the transportation of the agent from one computer system to another. The standardized form preserves the execution state of the encoded agent such that the agent can be decoded at the destination computer system and execution of the agent can resume by executing the instruction of the agent which sequentially follows operation "go". Each computer system of the network of the present invention can maintain an agent in whatever form is convenient for carrying out the computer instructions included in the agent. However, in transporting an agent from one computer system to another, each computer system of the network must use the standardized form of an agent during transportation much the same way that two people must agree to the vocabulary and syntax of a single language to communicate. The nature of the standardized form is discussed in greater detail in Appendices B, C and D, which are part of this disclosure.
As an agent can direct its own movement, i.e., transportation, through the network, means must be provided for specifying a number of parameters of a trip from a first computer system to a second computer system. A first parameter of a trip is the place to which the agent is transported, i.e., the destination of the trip. In one embodiment, a trip destination can be specified by name, by class, or by address. An address can be, for example, a location within a local area network.

Frequently, two or more pathways exist between a first computer system and a second computer system. One pathway may be quick but expensive and another may be time consuming but inexpensive. In such a case, no single 43a z14367~
pathway is always preferred in the transportation of an agent. Therefore, a second parameter of a trip is the "way" and the "means" for travel of an agent.
In addition, errors in configuring an agent for a trip to a nearby computer system can cause the agent to be inadvertently transported to distant computer systems at considerable expense to the creator of the agent.
Therefore, an important parameter of a trip is the amount of time or resources that can be expended in transporting the agent before the trip is aborted.
Yet another important consideration in controlling the transportation of agents throughout a network is security. For example, if all agents travelling to a particular computer system are granted ingress, the particular computer system may become overburdened with many processes requiring substantial processing, and throughput of the particular computer system can suffer.
Therefore, another parameter of a trip is the amount of resources the agent requires at the destination place.
Thus, the agent can be denied ingress to the place if the resources required by the agent are more than the place is able or willing to provide.
In one embodiment, a ticket 1306 (Figure 4A) controls the transportation of agent 150A and specifies place 220B
as the destination place. As shown in Figure 4A, agent 150A contains ticket 1306. Ticket 1306 defines the trip taken by agent 150A from place 220A to place 220B.
In one embodiment, ticket 1306 specifies (i) the destination place of the trip, (ii) the "way" or pathway and "means" agent 150A is to take to the destination place, (iii) the maximum amount of time within which the trip must be completed before the trip is aborted and (iv) the amount of resources agent 150A asks to use at the destination place. By specifying the amount of resources agent 150A is permitted to use at place 220B, place 220B
can determine whether agent 150A requires more resources WO 95/02219 ~ 14 3 6 '~ ~ PCTI(1S94/07397 than place 220B is configured to provide. In such a situation, place 220B can deny ingress to agent 150A.
Thus, using a ticket as described more completely below, an agent can completely define a pending trip between a first place and a second place.
As discussed above, the prior art process known to the inventors cannot manipulate objects of a class defined only in a previously occupied computer system. In other words, if a prior art process defines a class of objects and thereafter travels to a destination computer system, the process cannot manipulate objects of the class unless the destination computer system contains a definition of the identical class or the process defines the same class again within the destination computer system. As described in greater detail below, agents of the present invention, which travel from one place to another, transport definitions of classes which are not defined within the destination place. Thus, agents of the present invention can define a class within a first computer system, travel to a second computer system and manipulate objects of the class while executing within the second computer system. Agents of the present invention therefore represent a significant improvement over the known prior art processes.
What is also generally lacking in the prior art is a means to implement a complex, multi-level security system in a computer network in which computer processes are mobile. One type of security which is afforded by the present invention, and which is not suggested by the prior art, is the security afforded by places. As discussed in greater detail below, places, by granting or denying ingress to various agents, provide various levels of security and several places can be configured to provide multi-level, hierarchical security.
To assist in the understanding and visualization of the various aspects and embodiments of the present ~14'~~'~'~
invention, various representational and relational conventions are used in the drawings. Representational and relational conventions used herein are represented in Figures 5A, 5B and 5C. Computer system 120A (Figure 5A) is executing place 220A, place 220X, place 220Y, agent 150A, agent 150X and agent 150Y. Computer system 120A also contains objects 140A, 140B and 140X in a memory, e.g., mass memory or main memory (neither shown), of computer system 120A. Agent 150A and 150X occupy place 220A; place 220Y occupies place 220X; and agent 150Y
occupies place 220Y. Agent 150A owns objects 140A and 140B, and agent 150X owns object 140X. The relationships of occupancy and ownership are discussed in greater detail below. Briefly, occupancy of places provides various levels and types of security, and ownership determines which objects travel with an agent which performs operation "go".
Figure 5B is an alternative and equivalent representation of the various relationships represented in, and described above with respect to, Figure 5A. While Figure 5A accurately represents, e.g., that agent 150X
occupies place 220A and contains object 140X, the form of Figure 5A is less suitable for representing more complex relationships. Therefore, the tree structure of Figure 5B
is used in the drawings to represent more complex relationships in illustrating the various computer instructions of various embodiments of the present invention.
The tree structure of Figure 5B should not be confused with the class hierarchy tree which is shown in part in Figure 5C. Figure 5C represents the class relationships of various items represented in Figures 5A
and 5B. Objects 140A, 140B and 140X are members of class "Object" which is represented by class object 520.
Membership in a class is discussed in greater detail in the Glossary of Terms above and is represented by a dashed WO 95/02219 ~ ~ ~ , , PCT/US94/07397 line between the class and the object which is a member of the class. Class object 522 represents class "Process", which is a subclass of class "Object". Places 220X, 220Y
and 220A are members of class "Place", which is a subclass of class "Process" and is represented by class object 524.
Agents 150A, 150X and 150Y are members of class "Agent", which is a subclass of class "Process" and is represented by class object 526. Thus, places 220X, 220Y and 220A and agents 150A, 150X and 150Y are all members of class "Process" and are therefore computer processes. Objects 140A, 1408 and 140X are not members of class "Process" and are therefore not computer processes.
When an agent is transported from place to place, objects owned by the agent are transported with the agent from place to place. However, transporting an object can consume considerable resources if the object is large. In general, time in transporting an agent containing objects from a first place executing within a first computer system to a second place executing within a second computer system is substantially reduced by eliminating transportation of objects already at the second place.
This is easily demonstrated by considering a simple example. Agent 150A (Figure 6A) is executing within computer system 120A and occupies place 220A which is also executing within computer system 120A. Agent 150A owns objects 140A, 1408 and 140C. In this embodiment, object 1408 has digest 622. Digest 622 indicates that object 140B is interchangeable and that any object having a digest equal to digest 622 can be substituted for object 1408.
Place 2208 owns, and therefore contains, objects 624 and 626. Object 624 has digest 628. A process "contains"
all the objects owned by the process and the classes of which the process and the objects owned by the process are members.

21~'~6'~2 In performance of operation ~~go~~, as indicated above, agent 150A and all objects contained by agent 150A are represented in a standardized form. However, object 1408 is not included in the standardized form of agent 150A.
Instead, a copy of digest 622, i.e., digest 622-C
(Figure 6B), is included in the standardized form of agent 150A. Agent 150A, objects 140A and 140C, and digest 622-C are transported over communications link 102AB, in the direction of arrow A, to computer system 1208. Object 1408 is held within computer system 120A at least until it is determined that an equivalent interchangeable object is found within computer system 1208.
At computer system 1208 (Figure 6C), agent 150A, objects 140A and 140C, and digest 622-C are decoded. In decoding agent 150A, computer system 1208 recognizes digest 622-C and determines whether an object with an equivalent digest occupies place 2208. Object 624 occupying place 2208 has digest 628 which is equal to digest 622, and is therefore equal to digest 622-C. Since digest 628 is equal to digest 622, objects 1408 and 624 are interchangeable. Therefore, in decoding agent 150A, a copy, e.g., object 624-C, is made of object 624 and substituted within agent 150A for interchangeable object 1408. Agent 150A therefore owns object 624-C, which is a copy of object 624, in lieu of object 1408.
Thus, transportation of object 140B to place 2208 is obviated. Since an equivalent interchangeable object is found within computer system 1208, object 1408 can be deleted from computer system 120A. Of course, there are situations in which an equivalent interchanged object is not found on computer system 1208. In such a situation, object 1408 (Figure 6B) is retrieved from computer 120A as described more completely below.
An agent, occupying a first place, is capable of creating one or more clone processes of the agent and transporting each clone process to a respective place.
Thus, in effect, an agent is capable of traveling to and occupying several places simultaneously. Of course, it is not a single agent that travels simultaneously but rather the clones of the agent.
To travel to and occupy several places simultaneously, an agent issues an instruction which creates multiple clones of the agent and causes each clone to travel to a respective place. Each clone is an agent, and, at the time of a clone's creation, the clone is identical to the original agent, and therefore includes an execution state identical to that of the original agent.
For example, agent 150A, which occupies place 220A in computer 120A (Figure 7A), issues an instruction to computer system 120A which creates clones of agent 150A
and transports the clones to occupy one or more places, e.g., place 220B in computer system 120B and place 220C in computer system 120C, which are specified in the instruction issued. The instruction is called operation "send", and the issuance of the instruction by agent 150A
is called herein performance of operation "send" by agent 150A. Performance of operation "send" by agent 150A
as represented in Figures 7A, 7B, 7C and 7D is controlled, in this embodiment, by two tickets (not shown) within agent 150A. The two tickets control the transportation of respective clones and specify places 220B and 220C as the places to which respective clones of agent 150A are to be transported.
Agents 150A-1 and 150A-2 (Figure 7B) are clones formed from agent 150A, and are identical to agent 150A
except that agents 150A-1 and 150A-2 do not occupy place 220A. As described above, even the execution states of agents 150A-1 and 150A-2 are initially identical to the execution state of agent 150A. Agents 150A-1 and 150A-2 travel along intercomputer communications link 120ABC

~~1436'~ 2 (Figure 7C), as described in greater detail below, to respective places 220B and 220C.
After performance of operation "send" by agent 150A
(Figure 7D), agent 150A continues to occupy place 220A.
Agent 150A-1 occupies place 220B, and agent 150A-2 occupies place 220C.
Space in computer system 120A and time in transporting clones of agent 150A are saved by deferring complete cloning of an agent 150A so long as the travel path of one clone coincides with the travel path of another clone, i.e., at least two clones have some initial portion of their journey that is coextensive. For example, in Figure 7E, both clones of agent 150A must pass through computer system 120D to reach respective destination places 220B and 220C. Therefore, only a single clone of agent 150A, namely, agent 150A-1, is transported to computer system 120D. A second clone of agent 150A, namely, agent 150A-2, is formed in computer system 120D from agent 150A-1. Agents 150A-1 and 150A-2 are then transported to respective destination places 220B
and 220C. Thus, only a single clone of agent 150A is formed in computer system 120A and only a single clone is transported to computer system 120D saving space in computer system 120A and time in transporting agents 150A-1 and 150A-2 to their respective destinations.
Two agents exchange information by participating in a meeting between the two agents. In such a meeting, each agent is provided with a reference to the other agent. As discussed below, a first agent, which possesses a reference to a second agent, (i) can direct the second agent to take specific actions in accordance with instructions contained with the second agent and (ii) can transfer data to and receive data from the second agent.
In one aspect of the present invention, an agent is prevented from interfering with other agents by requiring that no agent be given a reference to a second agent ~I~3~~2 unless the second agent is given a reference to the first-mentioned agent. In that way, no agent can gain access to a second agent without granting reciprocal access.
A first agent, occupying a place, can initiate a meeting with a second agent occupying that place. During such a meeting, the first agent can transfer objects to and receive objects from the second agent, and the second agent can transfer objects to and receive objects from the first agent.
A subclass of class "Place" (Figure 5C) is class "Meeting Place" (not shown). In the illustrative example of Figures 8A-SF, place 220B is a member of class "Meeting Place" and is therefore a meeting place. Therefore, place 220B is sometimes referred to herein as "meeting place 220B". Meeting place 220B (Figure 8A), which is executing in computer system 120B, provides a means for agents 150A
and 150B, which occupy meeting place 220B, to communicate and share information in a meeting as indicated by arrows A and B. Petition 3106 within agent 150A defines and controls the meeting between agents 150A and 150B.
A meeting between agent 150A and 150B is arranged as illustrated by Figures 8B-8F. Agent 150A (Figure 8B) issues a first computer instruction, represented by arrow 851, directing meeting place 220B to arrange a meeting between agents 150A and 150B. The first instruction includes petition 3106, which controls the meeting, and which specifies that agent 150A wants to meet with agent 150B. On behalf of meeting place 220B
(Figure 8C), the engine issues a second computer instruction, represented by arrow 852, which notifies agent 150B that agent 150A has issued the first computer instruction, and which causes execution of a number of computer instructions within agent 150B to determine whether a meeting between agents 150A and 150B is acceptable.

Upon determination that the meeting between agents 150A and 150B is acceptable (Figure 8D), agent 150B
issues a reply, represented by arrow 853, to the second computer instruction, directing the engine, on behalf of meeting place 220B, to proceed in arranging the meeting between agents 150A and 150B. On behalf of meeting place 220B (Figure 8E), the engine conveys to agent 150A a reference to agent 150B, the conveyance being represented by arrow 854A, and conveys to agent 150B a reference to agent 150A, the conveyance being represented by arrow 854B. Agent 150A (Figure 8F) interacts with agent 150B using the reference to agent 150B, the reference being represented by arrow 855A, and agent 150B
interacts with agent 150A using the reference to agent 150A, the reference being represented by arrow 855B.
Two agents cannot participate in a meeting together unless the two agents occupy the same place. A hierarchy of places provides users of the present invention a mechanism to provide varying levels and types of access and security to various agents. The benefit of various levels and types of access and security is demonstrated by the illustrative example of Figures 9A-9E.
Figure 9A shows a building 902 having floors 902-1, 902-2, 902-3, 902-4 and 902-5. Floor 902-3 is a place within the place of building 902 and is therefore a subplace of building 902. Figure 9B shows floor 902-3 having rooms 902-3-1, 902-3-2, 902-3-3, and 902-3-4.
Room 902-3-2 is a place within the place of floor 902-3 and is therefore a subplace of floor 902-3.
The organization of building 902 lends itself to hierarchical security. For example, building 902 can be restricted to allow only people of a first level of security, e.g., military personnel, to enter. Floor 902-3 can be further restricted to allow only people of a second (higher) level of security, e.g., naval personnel, to enter. Room 902-3-2 can be still further restricted to 21~3~'~~
allow only people of a third (still higher) level of security, e.g., naval officers, to enter. Floor 902-4 can be restricted independently of floor 902-3, e.g., to allow only air force personnel to enter.
The engine in computer system 120X (Figure 9C) is executing places 220X1 and 220X2. Place 220X1 can represent, for example, building 902. Places 220X1-1 and 220X1-2, which represent, for example, floors 902-4-and 902-3, occupy place 220X1 (Figure 9D) and are therefore subplaces of place 220X1. Conversely, place 220X1 is a superplace of places 220X1-1 and 220X1-2. Place 220X1-2-1, which represents, for example, room 902-3-1, occupies place 220X1-2 (Figure 9E). Place 220X1-2-1 is therefore a subplace of place 220X1-2, and place 220X1-2 is a superplace of place 220X1-2-1.
In one embodiment of the present invention, two agents can only meet "face-to-face", i.e., when both agents occupy the same place. As used herein, an agent "occupies" one place only and does not simultaneously occupy subplaces or superplaces of the place. For example, an agent occupying place 220X1-2 does not simultaneously occupy place 220X1 or place 220X1-2-1.
Just as a first person in building 902 (Figure 9A) cannot communicate face-to-face with a second person on floor 902-3 unless the first person is also on floor 902-3, an agent occupying place 220X1 (Figure 9D) cannot communicate with an agent occupying place 220X1-2.
As a first person on floor 902-3 (Figure 9B) cannot communicate face-to-face with a second person in room 902-3-2 unless the first person is also in room 902-3-2, an agent occupying place 220X1-2 (Figure 9E) cannot communicate with an agent occupying place 220X1-2-1.
Such a place hierarchy enables a user of the present invention to restrict access to place 220X1, to further restrict access to place 220X1-2 and to restrict access to WO 95/02219 ~ PCT/US94/07397 _ 2 i ~ ~ 6'~ s~
place 220X1-2-1 further still. Additionally, the security hierarchy implemented is not necessarily directly related to the place hierarchy implemented. For example, access to place 220X1-2 can be more restrictive than access to a subplace of place 220X1-2, such as place 220X1-2-1. The restriction of access to a place is described in greater detail below. Thus, the place hierarchy enables a user of the present invention to construct sophisticated security hierarchies in which various agents are given various levels of access to other agents which occupy certain places, subplaces of those places, and subplaces of those subplaces.
To describe further details of the novel processes described above, a better understanding of the structure of the computer systems described above is required. In the embodiment shown in Figure 10, three computer systems are connected to form a computer network 100. Computer system 120A includes CPU 110A, network communications hardware 104A, and memory 117A. Input and output modules corresponding to input module 12 and output module 14 of Figure 1 are omitted for clarity. Additionally, mass memory 17A and main memory 17B of Figure 1 are combined to form memory 117A. Network communications hardware 104A
can be any device which enables CPU 110A to propagate signals across and to receive and interpret signals from the network.
Within memory 117A are a number of computer processes operating concurrently within CPU 110A. Network manager 130A is a computer process which coordinates data transmission between the various computer systems of the computer network. Operating system 131A is a computer process which coordinates the operation of the various components and resources of computer system 120A. For example, operating system 131A coordinates the use of components CPU 110A, memory 117A and I/O modules (not shown in Figure 10, see Figure 1). Engine 132A is a WO 95/02219 ~ ~ ~ PCT/US94/07397 computer process which interprets computer instructions of an object-oriented computer instruction set of this invention and processes information in the form of objects defined in that object-oriented computer instruction set.
For example, engine 132A effectuates concurrent execution of place 220A (Figure 4A) and agent 150A, both of which are processes constructed of the object-oriented computer instruction set of the present invention. Engine 132A is not shown in Figure 4A.
In addition to these processes operating within memory 117A (Figure 10) of computer system 120A, there are typically one or more processes which are user applications, e.g. user application 133A. User application 133A can request of engine 132A the creation of and/or manipulation of objects defined within the object-oriented instruction set interpreted by engine 132A.
Computer systems 1208 and 120C are configured similarly to computer system 120A. However, while the general structure of computer systems 120A, 1208 and 120C
are similar, computer systems 120A, 1208 and 120C can be otherwise heterogeneous.
The computer systems of computer network 100 are connected such that an agent process can be transported from one computer system to another. Computer systems 120A, 120B and 120C are connected to form a computer network by coupling the respective network communications hardware 104A, 1048 and 104C by means of communications links 102AB, 102BC and 102AC. Communications links 102AB, 102BC and 102AC can be any means by which data can be conveyed from one network communications hardware, e.g., network communications hardware 104A to another, e.g., network communications hardware 1048. For example, network communications link 102AB can be the public switched telephone network, in which case network communications hardware 104A and 1048 are modems and network managers 130A and 130B are capable of issuing commands to network communications hardware 104A and 104B, respectively, to establish and utilize communications via communications link 102AB i.e., the public switched telephone network.
Objects constructed of the object-oriented computer instruction set are executed by an engine, e.g., engine 132A (Figure 11A). Engine 132A has a communication infrastructure 132A-CI, a program portion 132A-P, and a data portion 132A-D. Data representing the state of the various objects executed by engine 132A, e.g., object 140A, are stored in data portion 132A-D of engine 132A. Data portion 132A-D of engine 132A is memory space within memory 117A
(Figure 10) reserved as work space for engine 132A and is generally inaccessible from the perspective of other processes on computer system 120A, e.g. operating system 131A and user application 133A.
As discussed above, an engine effectuates execution of processes and objects. Program portion 132A-P
(Figure 11A) of engine 132A includes computer instructions which effectuate execution of the objects represented in data portion 132A-D. The computer instructions which are combined to form program portion 132A-P can be of a known computer language. For example, in one embodiment, a program portion of an engine is constructed in accordance with the principles of the present invention and is constructed in accordance with the C++ programming language.
Communication infrastructure 132A-CI of engine 132A includes computer instructions which transport data between engines dispersed throughout network 100. Many aspects of the communication infrastructure of an engine are described in greater detail in Appendix D, which is a part of this disclosure.
56a ~~~3~'~~
Figure 11B is an alternative and equivalent representation of the computer network of Figure llA.
As indicated by Figure 11A, engine 132B of computer system 120B is configured in a manner that is directly analogous to the configuration of engine 132A.
Obiects within the Network As explained more completely below, places 220A and 220B (Figure 4A), agent 150A, ticket 1306, etc. are defined using "objects" according to the principles of this invention. As stated above, objects, which are formed according to the computer instruction set interpreted by engines 132A and 132B and which are interpreted by engine 132A, are stored in data portion 132A-D (Figure 11A). For example, object 140A in data portion 132A-D is formed according to the computer instruction set interpreted by engines 132A and 132B and is interpreted by engine 132A.
Brief Overview of the Computer Instruction Set of the Present Invention Each of the novel processes of this invention, and the features needed to define and support the processes, are defined by a set of computer instructions that are interpreted by an engine of this invention. The computer instructions of the present invention are object-oriented.
Therefore, data in the present invention, e.g., data which represent agent 150A, are organized into objects, each of which has an internal state and an external behavior. An object's properties define the object's internal state and the object's features define the object's external behavior (see Glossary of Terms above). Each object is an instance of a respective one of a number of classes.
According to the principles of this invention, all classes defined in the computer instruction set, except for mix-in classes which are described below, are subclasses of a class "Object", which is described more completely in Appendix A. Thus, each class that is described herein and which is not a mix-in class inherits the features and properties of class "Object".
In one embodiment, limited multiple inheritance is implemented using mix-in classes. "Mix-ins" or "mix-in classes" are classes which are not subclasses of class "Object". Examples of mix-in classes, which are described below and in Appendix A, include mix-in classes "Executed", "Named" and "Referenced". A non-mix-in class, alternatively called a "flavor" or a "flavor class", can be the immediate subclass of at most one flavor, but can be the immediate subclass of zero or more mix-in classes.
A mix-in class can be the immediate subclass of no class or of another mix-in class. Unless otherwise stated, a class is a flavor. No cycles in class hierarchy are permitted; i.e., no class is permitted to be both a subclass and a superclass of another class.
The use of mix-in classes allows features and properties to be defined once and used across a broad variety of classes. For example, mix-in class "Ordered"
defines operations for determining the relative order of two objects. Flavor classes "Association", "Citation", "List", "Identifier", "Pattern", "Permit", "Bit", "Boolean", "Character", "Number", "Octet" and "Time"
inherit from mix-in class "Ordered". Thus, associations, citations, lists, identifiers, patterns, permits, bits, booleans, characters, numbers, octets and times have an order relative to other members of the same class. For example, in the case of numbers, the number two is "after"
the number one and is "before" the number three.
The principal class of objects in the present invention is the class of processes. As described above, processes are either (i) places which are occupied by other processes, or (ii) agents which can (a) transport themselves from a first place, terminating occupancy of the first place, to occupy a second place and (b) interact with other agents occupying the same place.
Agent 150A (Figure 11A) is a process formed in accordance with the computer instruction set of the present invention whose execution is carried out by engine 132A. Agent 150A can (i) examine and modify itself, (ii) transport itself from a first place in network 100 (Figure 10) to a second place, and (iii) interact with other agents which occupy the second place.
Herein, a process is described as performing an operation that consumes arguments and produces zero or one result. However, a process, in and of itself, is an object that includes a collection of instructions, which are selected from the instructions described more completely below, and which is interpreted by an engine, e.g., engine 132A which is executing in CPU 110A. The performance of an operation by a process, or any other object of the disclosed instruction set, is in actuality the selection of and interpreting of a particular group of the instructions included in the process or object.
Interpreting the instructions of a process by an engine is herein alternatively called "interpreting the process".
Thus, when agent 150A performs an operation, agent 150A
provides instructions to engine 132A which in turn interprets the instructions and directs CPU 110A to carry out appropriate tasks effectuating performance of the operation by agent 150A. The interaction between a process and an engine is described more completely below.
An important aspect of the present invention is that the set of computer instructions described more completely below is implemented uniformly by the respective computer systems of network 100 (Figure 10). Computer systems 120A, 120B and 120C can use completely different and incompatible data structures, memory configurations, CPUs and operating systems. But, since an agent is capable of transporting itself to and from any of computer systems ~1436'~~
120A, 120B and 120C, it is important that each computer system implement the set of computer instructions of the present invention in a standardized fashion. As long as the computer instructions are uniformly implemented, computer systems 120A, 120B and 120C can be otherwise incompatible. Therefore, agents can travel freely among the computer systems of a heterogeneous, as well as a homogeneous, computer network.
Typically, prior art systems either required that processes travel only within homogeneous networks or that processes be specifically designed to execute computer instructions which were compatible with the computer system within which the process was executing. In the latter case, processes were configured for the particular system requirements of the computer system on which execution of the process began and for the particular system requirements of any computer systems to which the process was intended to travel. Thus, as the computer instructions of an agent of the present invention can execute on any computer system of a heterogeneous network, the present invention represents a substantial improvement over the prior art.
Agents As discussed above, the behavior of an agent is dependent in part on the internal state of the agent.
Therefore, prior to considering the external behavior of an agent in a network, several aspects of the internal state of an agent are briefly discussed. Each agent is a member of class "Agent".
It should be noted that no agents are instances of class "Agent," as class "Agent" is abstract. Class "Agent" is abstract as no implementation for operation "live" is defined or inherited by class "Agent". As discussed below and in Appendix A in greater detail, operation "live" defines the steps that are performed by a ~14~6?'~ ..
process, i.e., an agent or a place, upon creation of the process. These steps are collectively called the "central procedure" of the process. A central procedure for either agents or places is not provided as users of the present invention design and provide central procedures for agents and places to suit the particular needs of each user.
Thus, users of the present invention create concrete subclasses of class "Agent" and therein provide -implementations for operation "live".
Class "Agent" is a subclass of class "Process". (See Figure 5.) Class "Process" is a subclass of class "Object" and also inherits from mix-in class "Named".
Class "Object" inherits from mix-in class "Referenced".
An agent possesses the following attributes: (i) attributes "class" and "size" inherited from superclass "Object"; (ii) attribute "isProtected" inherited from mix-in class "Referenced"; (iii) attribute "name"
inherited from mix-in class "Named"; and (iv) attributes "brand", "permit" and "privateClasses" inherited from superclass "Process". Class "Agent" defines no attributes. Each of the above-mentioned attributes and classes is discussed in more detail in Appendix A.
Diagram 1270 (Figure 12) illustrates the class relations of the classes of which agent 150A is a member.
Agent 150A is a member of class "Agent" as agent 150A is shown to be contained within domain 1272 which represents class "Agent". Since class "Agent" is abstract, agent 150A is also a member of one or more subclasses (not shown) of class "Agent".
Domain 1272 is completely contained within domain 1274 which represents class "Process". Agent 150A
therefore inherits attributes "brand", "permit" and "privateClasses", which are defined by class "Process".
Additionally, all members of class "Agent" are also members of class "Process". Thus, as described above, class "Agent" is a subclass of class "Process".

Domain 1274 is completely contained within domain 1276 which represents class "Object". Agent 150A
therefore inherits attributes "class" and "size" which are defined by class "Object". Additionally, all members of class "Process", including members of class "Agent", are also members of class "Object". Thus, as described above, class "Process" is a subclass of class "Object".
Domain 1274 representing class "Process" is contained within domain 1278, which represents class "Named".
Connection 1278A shows that domain 1274 is contained within domain 1278 while accurately representing that domain~1276, representing class "Object", is not contained within domain 1278 and that domain 1278 is not contained within domain 1276. Class "Named", represented by domain 1278, is a mix-in class. As domain 1274 is contained within domain 1278, agent 150A is contained within domain 1278 and is therefore a member of mix-in class "Named". Mix-in class "Named" defines attribute "name" which is included in agent 150A.
Domain 1280 which represents mix-in class "Referenced" contains domain 1276, which represents class "Object", as indicated by connection 1280A. As domain 1276 is contained within domain 1280, agent 150A is contained within domain 1280 and is therefore a member of mix-in class "Referenced". Mix-in class "Referenced"
defines attribute "isProtected" which is included in agent 150A.
As every flavor of the disclosed embodiment of the present invention is a subclass of class "Object", every flavor is also a subclass of mix-in class "Referenced".
Thus, it is not necessary to separate the features of class "Referenced" from the features of class "Object".
However, the features of mix-in class "Referenced" are separated from the features of class "Object" to aid conceptualization and understanding of the present invention.

WO 95102219 ~ ~ ~ PCT/US94/07397 As discussed above in the Glossary of Terms, objects are identified within the disclosed computer instruction set by references. As discussed in greater detail below, in directing an object to perform an operation, the object is identified by a reference. Features defined by mix-in class "Referenced" operate on the reference which identifies the object. For example, attribute "isProtected" determines whether the reference to the object, and not the object itself, is protected. Features defined by class "Object" operate on the object identified by the reference. For example, attribute "class"
determines of which class the object is an instance.
It should be noted that no attribute is defined which provides information regarding the place occupied by agent 150A (Figure 12). However, the place occupied by agent 150A is maintained as a property of agent 150A.
Agent 150A can determine which place agent 150A occupies by execution of the "here" selector. The execution of selectors, and in particular the "here" selector, is discussed in greater detail in Appendix A. The property which is the place occupied by agent 150A is defined by class "Process". Thus, any process, i.e., either a place or an agent, includes a property which is the place occupied by the process.
While the only class disclosed herein and in Appendix A which is an immediate subclass of mix-in class "Named" is class "Process", subclasses of mix-in class "Named", which are not subclasses of class "Process", are defined by users of the present invention using the instructions disclosed herein and in Appendix A.
Therefore, class "Named" is a mix-in class.
Accents as Processes Associated with every process, both agents and places, is a central procedure which defines the primary behavior of the process as discussed above. The central procedure of a process is the method which implements an operation "live" as performed by the process. An engine initiates processing of a process by causing the process to perform operation "live", thereby causing execution of the process's central procedure. The provision of a method for an operation is discussed in greater detail below. When a process completes performance of operation "live", either successfully or otherwise, the process is terminated. The termination of a process is discussed in section 2.4.11 of Appendix A.
Mobility of Agents: Operation "Go"
As discussed above, an agent travels from one place to another by performance of operation "go".
Operation "go" is discussed in the context of the illustrative example of Figures 15A-15E. To travel from place 220A to place 220B, agent 150A (Figure 15A) performs operation "go". Figure 15A shows the state of network 1500 prior to performance of operation "go" by agent 150A.
Agent 150A owns objects 140A and 140B and occupies place 220A in engine 132A. Thus, agent 150A and place 220A
are processes which are simultaneously interpreted by engine 132A. In other words, place 220A and agent 150A are simultaneously performing operation "live". Engine 132B
interprets place 220B and agent 150B. Agent 150B occupies place 220B.
Communication infrastructure 132A-CI of engine 132A is connected to communication infrastructure 1322-CI of engine 1322 by communications link 102AZ. Engine 1322 interprets place 2202. Communication infrastructure 1322-CI
is also connected to communication infrastructure 132B-CI of engine 132B by communications link 102ZB. While in this embodiment only three computer systems are illustrated in network 1500, the number of computer systems in network 1500 is an arbitrary number. Hence, 64a the computer systems illustrated in Figure l5A.are illustrative of the principles of the invention and are not intended to limit the invention to the particular network illustrated.
Performance of operation "go" by agent 150A, in this embodiment, requires transportation of agent 150A from engine 132A through engine 1322 to engine 132B. Hence, operation "go" requires action on the part of engine 132A, i.e., the source engine, engine 1322, i.e. the transit engine, and engine 132B, i.e., the destination engine.
Operation "Go" as Performed by the Enaine Processing the Source Place Figure 13A shows a portion of the internal state, including a portion of the execution state, of agent 150A
immediately prior to performance of operation "go" by agent 150A. The execution state of a process is described in greater detail below. Stack 1304 is a part of the execution state of agent 150A. Stack 1304 contains, at top "T", ticket 1306. Ticket 1306 is described more completely below. Arguments consumed by performance of operation "go" are popped from stack 1304 and stack 1304 is therefore the "current stack" in the context of operation "go". Place 220A is a property of agent 150A, indicating that agent 150A occupies place 220A.
Figure 13B shows a portion of the internal state, including a portion of the execution state, of agent 150A
immediately following performance of operation "go" by agent 150A. Figure 13B is described in greater detail below.
Logic flow diagram 1400 (Figure 14A) illustrates operation "go" as carried out by engine 132A (Figure 15A).
Engine 132A is the source engine as agent 150A is executing within engine 132A when performance of operation "go" is initiated. The steps of logic flow diagram 1400 ~~~36~2 (Figure 14A) are discussed in the context of the trip from place 220A (Figure 15A) to place 220B taken by agent 150A.
In performance of operation "go" by agent 150A, engine 132A (Figure 15A) determines whether agent 150A is the agent requesting performance of operation "go" in an access test step 1402 (Figure 14A). If operation "go" is requested by an object other than agent 150A (Figure 15A), processing transfers from access test step 1402 (Figure 14A) to terminal step 1404 in which an exception of class "Process Not Current" is thrown causing operation "go" to fail. If, however, operation "go" is requested by agent 150A (Figure 15A), processing transfers from access test step 1402 (Figure 14A) to a "canto" test step 1406.
In "canto" test step 1406, engine 132A (Figure 15A) determines whether agent 150A is permitted to perform operation "go". In "canto" test step 1406, engine 132A
(Figure 15A), queries attribute "permit" of agent 150A.
Attribute "permit" as discussed above, is one of a plurality of attributes of agent 150A. The various attributes of agent 150A are described in greater detail in Appendix A in conjunction with class "Process". The query of attribute '!permit" of agent 150A produces property "permit" of agent 150A. Specifically, permit 1612 (Figure 16) is property "permit" of agent 150A and is therefore produced by querying attribute "permit" of agent 150A. In this embodiment, property "permit" is one of a number of properties of agent 150A.
Permit 1612 (Figure 17) includes, among several properties which are described in greater detail in Appendix A, properties "charges", "canto", and "age" which are integer 1702, boolean 1704, and integer 1706, respectively. A boolean is an object having one of only two possible values: "true" or "false". In "canto" test step 1406 (Figure 14A), engine 132A (Figure 15A) queries attribute "canto" of permit 1612 (Figure 17), thereby ~1~ ~36'~2 s .
producing boolean 1704 which is property "canto" of permit 1612, and compares boolean 1704 to "true".
If boolean 1704 has a value of "false", processing transfers from "canto" test step 1406 (Figure 14A) to terminal step 1408. Agent 150A is not permitted to travel by performance of operation "go" and the operation fails and throws an exception that is a member of class "Permit Violated" in terminal step 1408. Exceptions and their classification are described more completely in Appendix A.
Otherwise, if boolean 1704 (Figure 17) has a value of "true", processing transfers from "canto" test step 1406 (Figure 14A) to an effectuate move step 1410. Effectuate move step 1410 is described in greater detail below in conjunction with Figure 14B. Processing transfers from effectuate move step 1410 (Figure 14A) to terminal step 1412 in which operation "go", to the extent engine 132A is involved, completes successfully.
Effectuate move step 1410, which is performed by engine 132A (Figure 15A) as described above, is represented by logic flow diagram 1410 (Figure 14B). In the first step of logic flow diagram 1410, i.e., route agent step 1414, engine 132A (Figure 15A) selects an engine to use in transporting agent 150A to the trip destination, i.e., a "transfer destination". As used herein, the transfer destination is the next engine used in the trip and should not be confused with the "trip destination". Ticket 1306 (Figure 13A) defines a place as the destination of a trip; therefore, the "trip destination" is a place. In the course of the trip defined by ticket 1306 (Figure 13A), agent 150A (Figure 15A) is transferred from one engine to another.
Therefore, a "transfer destination" is an engine. In general, a trip can require several transfers of an agent before the agent reaches an engine which contains a trip destination.

In step 1414 (Figure 14B), a way object, i.e., a member of class "Way", which defines a transfer destination is produced. The transfer destination can be, for example, (i) the engine which currently contains agent 150A, i.e., engine 132A, in which case agent 150A is not transferred, (ii) another engine in the region which contains engine 132A, or (iii) an engine in another region. In either of possibilities (ii) or (iii),-agent 150A is transferred to the transfer destination as described below.
The term "region" is defined above in the Glossary of Terms. The grouping of engines into regions is significant in the routing of an agent as each engine typically contains more detailed information regarding other engines within the same region than engines of other regions. This is true because engines of a single region are configured by a single person or organization, which is therefore called the "provider" of the region.
Therefore, each engine of a region can be given detailed information regarding other engines of the region as all such engines are provided by a single provider.
The grouping of engines into regions is helpful in routing agents in large and complex networks. When an agent travels between regions, it is not imperative that the source engine contain any information regarding the transfer destination engine. It is sufficient that the source engine can determine to which region the agent is traveling and can transfer the agent to an engine in that region. Once transferred to an engine in the region containing the place that is the trip destination, the agent can be more easily routed to an engine containing that place.
Logic flow diagram 1414 (Figure 14C) shows the steps carried out in route agent step 1414 (Figure 14B) and is discussed below in greater detail.

Processing transfers from route agent step 1414 to an isolate agent step 1416 where agent 150A (Figure 15A) is isolated. Isolation of a process is described in greater detail in Appendix A. Briefly, agent 150A is isolated (i) by voiding all references within agent 150A to other processes, and to objects owned by other processes, and (ii) by voiding within all other processes references to agent 150A and to objects owned by agent 150A. Processing transfers from isolate agent step 1416 (Figure 14B) to existing step 1418.

WO 95/0221 ~ ~ 3 ~ ~ PCT/US94/07397 In exiting step 1418, place 220A (Figure 15A) notes the departure of agent 150A by performing operation "exiting" at the request of engine 132A. Operation "exiting" is described below in greater detail.
Processing transfers from exiting step 1418 to a next hop is here test step 1420.
In next hop is here test step 1420, engine 132A
(Figure 15A) compares the transfer destination determined in route agent step 1414 (Figure 14B) to the current engine, i.e., engine 132A (Figure 15A). If the transfer destination is the current engine, i.e., engine 132A, processing transfers from next hop is here test step 1420 (Figure 14B) to deliver agent step 1422. In the illustrative example of Figures 15A-15E, agent 150A is transferred to engine 1322, as discussed below; thus, the transfer destination is not the current engine, i.e., engine 132A, and processing does not transfer to deliver agent step 1422 (Figure 14B). However, deliver agent step 1422 and the steps which follow are described for completeness.
In deliver agent step 1422, agent 150A (Figure 15A) is delivered to a place within engine 132A which satisfies ticket 1306 (Figure 13A). Deliver agent step 1422 (Figure 14B) is represented by logic flow diagram 1422 (Figure 14E) and is discussed in greater detail below. Processing transfers from deliver agent step 1422 (Figure 14B) to a first exception test step 1424 in which engine 132A
(Figure 15A) determines whether delivery of agent 150A
threw an exception. If no exception is thrown in deliver agent step 1422 (Figure 14B), processing transfers from first exception test step 1424 to terminal step 1428.
Terminal step 1428 completes move step 1410 and so processing transfers to terminal step 1412 (Figure 14A) which is described above.
If, on the other hand, an exception is thrown in deliver agent step 1422 (Figure 14B), processing transfers from first exception test step 1424 to step 1426 in which agent 150A is delivered to purgatory according to logic flow diagram 1422 (Figure 14E). Purgatory is a place within each engine and which never denies ingress to a process. In entering purgatory, the local permit of agent 150A can be severely restricted. For example, property "canto" of the local permit is set to "true" and properties "canCharge", "canCreate" "canDeny" "canGrant" "canRestart" " "
, canSend and "canTerminate" are set to "false". Properties "charges"
and "age" are set just large enough that agent 150A can detect and analyze the exception that sent agent 150A to purgatory, and can travel to another place by performance of operation "go". The various properties of a permit are described in greater detail in Appendix A.
Processing transfers from step 1426 to terminal step 1428 and so effectuate move step 1410 (Figure 14A), completes successfully, as described above.
As discussed above, if the transfer destination determined in route agent step 1414 (Figure 14B) is the current engine, i.e., engine 132A, processing transfers from next hop is here test step 1420 (Figure 14B) to deliver agent step 1422. Conversely, if the transfer destination is not the current engine, processing transfers from next hop is here test step 1420 to a form destination step 1430. In form destination step 1430, a destination object which specifies the transfer destination is formed. Processing transfers from form destination step 1430 to encode agent step 1432.
In encode agent step 1432, engine 132A (Figure 15A) encodes agent 150A according to the encoding rules of the Telescript Encoding Rules, included as Appendix B, which is a part of this disclosure. Encoding agent 150A results in (i) agent 150A, (ii) all objects owned by agent 150A
including objects 140A and 140B and (iii) the classes of which agent 150A and all objects owned by agent 150A are members being represented in encoded agent 150A-E (Figure 15B) in a standardized binary form. The destination object formed in form destination step 1430 (Figure 14B), i.e., destination object 150A-E-D (Figure 21), is included in encoded agent 150A-E and is discussed in greater detail below. As shown in Figure 15B, encoded agent 150A-E is stored in communication infrastructure 132A-CI of engine 132A. Agent 150A (Figure 15A) is retained by engine 132A
until the transfer of encoded agent 150A-E is complete as described below in greater detail.
Processing transfers from encode agent step 1432 to transfer out step 1434. In transfer out step 1434, engine 132A (Figure 15B) initiates transfer of encoded agent 150A-E according to the destination object formed in form destination step 1430 (Figure 14B). In this example, the destination object specifies engine 1322 (Figure 15B) as the transfer destination of encoded agent 150A-E. Encoded agent 150A-E is transferred as ordinary binary data from communications infrastructure 132A-CI, across communications link 102AZ, to communications infrastructure 1322-CI of engine 1322. The transportation of data between communications infrastructures 132A-CI and 1322-CI is described in greater detail in Appendix C, which is a part of this disclosure. As discussed above, the transfer of encoded agent 150A-E is initiated in transfer agent out step 1434. Processing according to logic flow diagram 1410 (Figure 14B) by program portion 132A-P of engine 132A
proceeds while the transfer of encoded agent 150A-E between communications infrastructures 132A-CI and 1322-CI
continues.
Processing transfers from transfer out step 1434 to a second exception test step 1436 in which engine 132A
determines whether transfer out step 1434 threw an 72a WO 95/02219 ~ ~ PCTIUS94/07397 exception, i.e., whether initiation of the transfer of encoded agent 150A-E failed. If transfer out step 1434 fails, processing transfers from second exception test step 1426 in which agent 150A is delivered to purgatory as described above. Conversely, if transfer out step 1434 succeeds, processing transfers from second exception test step 1436 to step 1438.
In step 1438, engine 132A (Figure 15B) adds agent 150A to a list of pending transfers. Agent 132A is retained in the list of pending transfers during the transfer of agent 150A. When the transfer is complete, agent 150A is removed from the list of pending transfers and discarded by engine 132A. If the permit of agent 150A
expires while agent 150A is still on the list of pending transfers, i.e., if the actual age of agent 150A reaches property "age" of the effective permit of agent 150A, the transfer of encoded agent 150A-E is aborted; agent 150A is removed from the list of pending transfers; and agent 150A
throws an exception of class "Permit Expired", thereby failing to perform operation "go" successfully. The "effective permit" of a process is described in greater detail in Appendix A. Processing transfers from step 1438 to terminal step 1428 in which logic flow diagram 1410 (Figure 14B), and therefore effectuate move step 1410 (Figure 14A), completes.
As described above, in route agent step 1414, engine 132A (Figure 15A) determines the transfer destination for agent 150A. Logic flow diagram 1414 (Figure 14C) shows the steps carried out in route agent step 1414 (Figure 14B).
The following discussion of logic flow diagram 1414 (Figures 14C and 14D) considers circumstances beyond the scope of the illustrative example of Figures 15A-15E.
Logic flow diagram 1414 (Figures 14C and 14D) is therefore described independently of the illustrative example of Figures 15A-15E. To provide a framework for discussion, ~143~72 however, logic flow diagram 1414 (Figures 14C and 14D) is described in the context of agent 150A (Figure 15A) performing operation "go" to effectuate a trip which originates in place 220A within engine 132A. The discussion of logic flow diagram 1414 (Figures 14C and 14D) is not limited to (i) a trip in which agent 150A is transferred to an engine other than engine 132A (Figure 15A) or (ii) a network in which engines 132A, 1322 and 132B are of the same region. Furthermore, the following discussion does not limit ticket 1306 (Figure 13A), which defines the trip taken by agent 150A (Figure 15A), to either including or excluding specification of a way, a means, a telename, or a provider. Ticket 1306 (Figure 13A), in the context of logic flow diagram 1414 (Figures 14C and 14D), can define a trip (i) to a place within engine 132A (Figure 15A), (ii) to a place within an engine in the same region as engine 132A, or (iii) to a place within an engine in a region different than the region which contains engine 132A.
Recall that for operation "go", ticket 1306, which is consumed as an argument to operation "go", specifies the destination and other characteristics of the trip. In logic flow diagram 1414, engine 132A (Figure 15A) in test step 1440 queries attribute "way" of ticket 1306 (Figures 13 and 18A), thereby producing property "way" of ticket 1306. Property "way" of ticket 1306 is way 1820 (Figure 18A). Property "way" of ticket 1306 is optional and can therefore alternatively be a nil (not shown).
If property "way" of ticket 1306 is a nil, processing transfers from test step 1440 to a ticket has provider test step 1460 which is described below in greater detail;
Conversely, if property "way" of ticket 1306 is a way, i.e., is not a nil, processing transfers from test step 1440 to step 1442 in which engine 132A (Figure 15A) queries attribute "name" of way 1820 (Figure 18A), thereby producing the telename that is property "name" of way ~I~3~72 1820. Also in step 1442 (Figure 14C), engine 132A (Figure 15A) produces and copies property "authority" of that telename. The resulting "authority", which is represented by an octet string as described in Appendix A, is used to determine the transfer destination, as described below.
Processing transfers from step 1442 to a way has means test step 1444 (Figure 14C) in which attribute "means" of way 1820 (Figure 18B) is queried, thereby producing property "means" of way 1820, which is means object 1822. Property "means" is optional and can therefore alternatively be a nil (not shown).
A means object, i.e., a member of class "Means", is an object which specifies an engine by specifying (i) an intercomputer communications medium and (ii) transfer instructions to reach a specific engine via that medium.
For example, a means object can specify the public switched telephone network (PSTN) and specify modem instructions which include the telephone number of a specific modem through which to transfer the agent to a destination engine. As a means object contains all information needed to route agent 150A to a destination engine, other properties of ticket 1306 are ignored if ticket 1306 contains a means object.
Therefore, if property "means" of way 1820 is not nil, i.e., is means object 1822 (Figure 18B), processing transfers from way has means test step 1444 (Figure 14C) to step 1446. In step 1446, way 1820 is designated as the way which defines the transfer destination. It should be noted that means object 1822 defines the engine that is the destination of the first hop, i.e., the first transfer of agent 150A (Figure 15A), and not the destination of the trip defined by ticket 1306 (Figure 18A), which is necessarily a place rather than an engine. Processing transfers from step 1446 to terminal step 1448 in which the processing of logic flow diagram 1414, i.e., route agent step 1414 (Figure 14B), completes successfully.

WO 95/02219 . PCTIUS94/07397 ~1436'~2 Conversely, if property "means" of way 1820 (Figure 18B) is nil, way 1820 contains no means object and processing transfers from way has means test step 1444 (Figure 14C) to a ticket has name test step 1450. In ticket has name test step 1450, engine 132A (Figure 15A) produces property "destinationName" of ticket 1306 (Figure 18A), which is telename 1818, and compares telename 1818 to a nil. As property "destinationName" of ticket 1306 is optional, property "destinationName" of ticket 1306 can be a nil (not shown).
If property "destinationName" of ticket 1306 is a nil, processing transfers from ticket has name test step 1450 to ticket has provider test step 1460, which is described below. Conversely, if property "destinationName" of ticket 1306 is telename 1818, processing transfers from ticket has name test step 1450 to step 1452. In step 1452, engine 132A (Figure 15A) produces a way object which defines a transfer path for agent 150A to the transfer destination. Engine 132A
consults a finder, which is described in greater detail below, to produce a way object to a region associated with the authority produced and copied in step 1442 (Figure 14C) above, and a place within that authority whose name is equal to telename 1818 (Figure 18A). The production of such a way object by use of a finder is described in greater detail below.
Processing transfers from step 1452 to a way out is here test step 1454 in which engine 132A (Figure 15A) determines whether the way object produced in step 1452 (Figure 14C) defines a transfer to engine 132A (Figure 15A). If the produced way object does not define a transfer to engine 132A, processing transfers from way out is here test step 1454 to terminal step 1456 in which the processing of logic flow diagram 1414, i.e., route agent step 1414 (Figure 14B), completes successfully.

Conversely, if the produced way object defines a transfer to engine 132A (Figure 15A), processing transfers from way out is here test step 1454 to step 1458 in which engine 132A sets a flag indicating that agent 150A is to be transferred to engine 132A. Processing transfers from step 1458 to a second ticket has name test step 1480 which is described below in greater detail.
As discussed above, if property "way" of ticket 1306 is nil, processing transfers from ticket has way test step 1440 to ticket has provider test step 1460. In ticket has provider test step 1460, engine 132A produces property "destinationAddress" of ticket 1306, which is a teleaddress 1814 (Figure 18A), and produces property "provider" of teleaddress 1814. If either property "destinationAddress" of ticket 1306, or property "provider" of teleaddress 1814, is a nil, processing transfers from ticket has provider test step 1460 to second ticket has name test step 1480, which is described below in greater detail. Conversely, if property "destinationAddress" of ticket 1306 (Figure 18A) and property "provider" of the teleaddress 1814 are not nil, ticket 1306 specifies a region which contains the trip destination and processing transfers from ticket has provider test step 1460 to step 1462.
In step 1462, engine 132A (Figure 15A) consults a finder, as described below in greater detail, to produce a way object which defines a transfer to an engine of a region denoted by property "provider" of property "destinationAddress" of ticket 1306. Processing transfers from step 1462 (Figure 14C) to a first found test step 1464 in which engine 132A (Figure 15A) determines whether a way object is successfully produced in step 1462. If a way object is produced in step 1462, processing transfers from first found test step 1464 to a second way out is here test step 1472, which is described in greater detail below. Conversely, if no way object is produced in step _77_ 1462, processing transfers from first found test step to step 1466.
In step 1466, engine 132A (Figure 15A) consults a finder, as described below in greater detail, to produce a way object which defines a transfer to a region denoted by an item of property "routingAdvice" of the teleaddress that is property "destinationAddress" of ticket 1306. The properties of a teleaddress, and properties "provider" and "routingAdvice" in particular, are described in greater detail in Appendix A. Briefly, the items of property "routingAdvice" of a teleaddress denote providers of zero or more regions which are believed by the creator of the teleaddress to be capable of transferring agent 150A to the trip destination defined by ticket 1306 (Figure 18A).
Processing transfers from step 1466 (Figure 14C) to a second found test step 1468 in which engine 132A (Figure 15A) determines whether a way object is successfully produced in step 1466. If no way object is produced in step 1466, processing transfers from second found test step 1468 to terminal step 1470 in which an exception of class "Destination Unavailable" is thrown. Conversely, if a way object is produced in step 1466, processing transfers from second found test step 1468 to second way out is here test step 1472.
In second way out is here test step 1472, engine 132A (Figure 15A) determines whether the way object produced in either step 1462 or step 1466 (Figure 14C) defines a transfer to engine 132A (Figure 15A). If the produced way object does not define a transfer to engine 132A, processing transfers from second way out is here test step 1472 to terminal step 1474 in which the processing of logic flow diagram 1414, i.e., route agent step 1414 (Figure 14B), completes successfully.
Conversely, if the produced way object defines a transfer to engine 132A (Figure 15A), processing transfers 78a _ WO 95102219 PCT/US94107397 ~'1~ ~3~'~2 from second way out is here test step 1472 to step 1476 in which engine 132A (Figure 15A) derives a telename, which identifies the destination of the trip defined by ticket 1306, from property "location" of the teleaddress that is property "destinationAddress" of ticket 1306.
Engine 132A can derive the telename in any way which is convenient and efficient. For example, engine 132A can use a table such as dictionary 2000 (Figure 20A) whose keys are octet strings, e.g., octet string 2000K1, and whose values are telenames, e.g., telename 2000V1. For example, if property "location" of teleaddress 1814 (Figure 18A) equals octet string 2000K1 (Figure 20A), telename 2000V1 is produced by engine 132A as the telename that specifies a place that is the destination of the trip defined by ticket 1306 (Figure 18A).
Addresses are assigned to locations within a region according to an addressing scheme designed and implemented by the designer and implementer of the engines of the region. The disclosed instruction set does not prescribe a particular addressing scheme. Each region is free to implement the most efficient and convenient addressing scheme for that particular region. Thus, as each sovereign state in the physical world designs and implements its own addressing scheme in providing a postal service, each region is free to implement a unique addressing scheme for that region. Of course, the addressing schemes used within two or more regions need not be unique.
In implementing an addressing scheme, engines of a region assign teleaddresses to locations which include one or more places. In assigning a teleaddress, an engine produces and stores a telename which specifies one or more of the places within the location specified by the teleaddress. In other words, an engine, which assigns teleaddresses, maintains information which links teleaddresses to telenames of places to which respective _79_ WO 95/02219 - PCTlUS94/07397 ~1436'~2 teleaddresses are assigned. In this way, an engine, e.g., engine 132A (Figure 15A), can produce a telename of a place having a specified teleaddress. The particular mechanism by which such a telename is stored and produced is left up to the person or organization which designs and implements the engines of a given region.
Processing transfers from step 1476 to step 1478 in which engine 132A (Figure 15A) sets a flag indicating that agent 150A is to be transferred to an engine within the region which contains engine 132A. Processing transfers from step 1478 (Figure 14C) to second ticket has name test step 1480 (Figure 14D). Processing also transfers to second ticket has name test step 1480 from step 1458 (Figure 14C) and from ticket has provider test step 1460, both of which are described above in greater detail.
In second ticket has name test step 1480 (Figure 14D), engine 132A (Figure 15A) determines whether property "destinationName" of ticket 1306 (Figure 18A), is a nil or telename 1818. If property "destinationName" of ticket 1306 is telename 1818, processing transfers from second ticket has name test step 1480 (Figure 14D) to step 1482 in which engine 132A (Figure 15A) produces a copy of telename 1818 (Figure 18A) and the copy supersedes the telename derived in step 1476 (Figure 14C) as the telename of the transfer destination. Processing transfers from step 1482 (Figure 14D) to step 1483. Additionally, if property "destinationName" of ticket 1306 is a nil, processing transfers from second ticket has name test step 1480 directly to step 1483.
In step 1483, engine 132A (Figure 15A) consults a finder, as described below in greater detail, to produce a way object which defines a transfer to a place denoted by the telename produced in either step 1476 (Figure 14C) or step 1482 (Figure 14D) as described above. Processing transfers from step 1483 to test step 1484 in which engine 132A (Figure 15A) determines whether step 1458 (Figure ~~~~s~~
14C) sets the flag which indicates that the destination of the transfer is engine 132A (Figure 15A). If the flag is not set, processing transfers from test step 1484 to test step 1490 which is described below. Conversely, if the flag is set, the transfer of agent 150A (Figure 15A) should be to a place within engine 132A and processing transfers from test step 1484 (Figure 14D) to a third way out is here test step 1486.
In third way out is here test step 1486, engine 132A
(Figure 15A) determines whether the way object produced in step 1483 (Figure 14D) defines a transfer to engine 132A
(Figure 15A). If the produced way object does not define a transfer to engine 132A, processing transfers from third way out is here test step 1486 (Figure 14D) to terminal step 1488 in which an exception of class "Destination Unavailable" is thrown and processing according to logic flow diagram 1414 (Figures 14C and 14D), i.e., route agent step 1414 (Figure 14B), completes. Conversely, if the produced way object defines a transfer to a place within engine 132A (Figure 15A), processing transfers from test step 1486 (Figure 14D) to test step 1490.
In test step 1490, engine 132A (Figure 15A) determines whether step 1478 (Figure 14C) sets the flag which indicates that the destination of the transfer is the region which includes engine 132A (Figure 15A). If this flag is not set, processing transfers from test step 1490 to terminal step 1496 in which processing according to logic flow diagram 1414 (Figures 14C and 14D), i.e., route agent step 1414 (Figure 14B), completes successfully. Conversely, if the flag is set, the transfer of agent 150A (Figure 15A) should be to a place within the region which includes engine 132A and processing transfers from test step 1490 (Figure 14D) to way out is this region test step 1492. In way out is this region test step 1492, engine 132A (Figure 15A) determines whether the way object produced in step 1483 WO 95/02219 ' PCT/US94/07397 ~n3s~2 (Figure 14D) defines a transfer to a place within the region which includes engine 132A (Figure 15A). If the produced way object does not define a transfer to a place within the region which includes engine 132A, processing transfers from way out is this region test step 1492 (Figure 14D) to terminal step 1494 in which an exception of class "Destination Unavailable" is thrown and processing according to logic flow diagram 1414 (Figures 14C and 14D), i.e., route agent step 1414 (Figure 14B), completes.
As discussed above with respect to steps 1452, 1462, 1466 and 1483 (Figure 14C), engine 132A (Figure 132A) uses a finder to route agent 150A in a transfer toward a destination of the trip defined by ticket 1306 (Figure 13A). A finder is used to determine the transfer destination for an agent. For example, ticket 1306 (Figure 13A) defines place 220B (Figure 15A) within engine 132B as a destination of the trip. However, to reach engine 132B, agent 150A must first be transferred to engine 1322, which is therefore the transfer destination.
Finder 2050 (Figure 20B) is used to produce a telename of an engine that is a transfer destination toward a place, which is identified by a second telename. In one embodiment, the produced telename identifies an engine by identifying the engine place processed by the engine. The second telename can identify a specific place or an authority, thereby specifying all places of the identified authority. Telenames are described in greater detail in Appendix A.
In the context of step 1452 (Figure 14C), the second telename is telename 1818 (Figure 18A), which is property "destinationName" of ticket 1306. If finder 2050 (Figure 20B) contains no information regarding telename 1818, the second telename, in the context of step 1452 (Figure 14C), is a telename which identifies the authority produced and copied in step 1442 as described above. In the context of _ WO 95102219 PCTIUS94107397 ~1~3G'~2 step 1462, the second telename identifies the authority specified in property "provider" of teleaddress 1814 (Figure 18A) of ticket 1306. In the context of step 1466 (Figure 14C), the second telename identifies the authority specified in an item of the list that is property "routingAdvice" of telename 1814 (Figure 18A) of ticket 1306. In the context of step 1483 (Figure 14D), the second telename is either telename 1818 (Figure 18A) of ticket 1306, if property "destinationName" of ticket 1306 is not nil, or the telename produced in step 1476 (Figure 14C), otherwise. Finder 2050 (Figure 20B) is described below in the context of step 1483 (Figure 14D) in which the second telename is telename 1818 (Figure 18A) of ticket 1306.
In each of steps 1452 (Figure 14C), 1462, 1466, and 1483 (Figure 14D), the telename produced by use of finder 2050 (Figure 20B) is used to form a way object which defines a transfer to the engine identified by the produced telename.
Engine 132A determines to which engine to transfer agent 150A, i.e., the "transfer destination", by use of a finder 2050 (Figure 20B). Finder 2050 can be any data structure which allows engine 132A to determine, from a trip destination, a transfer destination that moves agent 150A toward the trip destination. In one embodiment, finder 2050 is a dictionary whose keys are telenames and whose values are also telenames. The keys, e.g., telenames 2050K1-2050K6, specify authorities. The values, e.g., telenames 2050V1-2050V6, identify engines through which places identified by each respective corresponding authority can be reached. If telename 2050K1 (Figure 20B) of finder 2050 has the same authority as that of telename 1818, telename 2050V1 identifies engine 1322 as the transfer destination.
In one embodiment of the present invention, the keys of finder 2050 include a nil 2050K7, which is associated ~~43~'~~
with a telename 2050V7. Telename 2050V7 identifies an engine which contains information regarding a substantial portion of network 1500 (Figure 15A) and networks to which network 1500 can be connected, either directly or indirectly. The engine which is identified by telename 2050V7 is therefore more likely to have success in routing agent 150A toward a trip destination which satisfies ticket 1306 (Figure 18A). If no telename which is a key of finder 2050 (Figure 20B) is of the same authority as telename 1818, telename 2050V7 (Figure 20B), which is associated with nil 2050K7, identifies the transfer destination. The following example is illustrative.
Suppose, for example, that engine 132A is executing within a small personal computer system and therefore interprets places of only one authority. Suppose further that engine 1322 (Figure 15A) is executing within a large, multi-user, mainframe computer system which interprets places of many separate authorities and to which many agents of many different authorities travel. The finder of engine 132A would, in such a case, be very small and provide little information regarding the transportation of agents to places of particular authorities. However, the finder of engine 1322 is likely to be much more extensive and comprehensive, and therefore much more likely to provide information regarding the routing of a particular agent to places of particular authorities. In such a case, the finder of engine 132A associates with a nil a telename which identifies engine 1322 as the transfer destination when the finder of engine 132A contains no information regarding the authority of the trip destination place.
If no transfer destination is successfully determined from consulting finder 2050 (Figure 20B), an engine can implement any of the following or other policies according to the implementation chosen by the provider of the engine. If ticket 1306 (Figure 18A) contains citation __ WO 95/02219 PCT/US94107397 z~~3s~~
1816, engine 132A (Figure 15A) can (i) produce a telename to the current place of agent 150A and attempt to find a place of the cited class in the steps described below, (ii) throw an exception and produce no telename, thereby rejecting ticket 1306 (Figure 18A) as vague, or (iii) produce a telename of a place which is designated as a destination of vaguely specified trips. Similarly, if ticket 1306 includes no citation, engine 132A (Figure 15A) can (i) throw an exception and produce no telename, thereby rejecting ticket 1306 (Figure 18A) as vague, or (ii) produce a telename of a place which is designated as a destination of vaguely specified trips.
In the example of Figures 15A-15E, engine 132A
(Figure 15A) consults finder 2050 (Figure 20A) and determines that, to reach place 220B (Figure 15A) of engine 132B, agent 150A is to be transferred to engine 1322. Therefore, the engine place of engine 1322 is the transfer destination of agent 150A and a way object, which defines a transfer of agent 150A to engine 1322, is produced.
Thus, processing according to logic flow diagram 1414 (Figures 14C and 14D) produces a way object which defines a transfer of agent 150A (Figure 15A) to move agent 150A
toward the destination of the trip defined by ticket 1306.
As discussed above in the context of logic flow diagram 1410 (Figure 14B), agent 150A (Figure 15A) is delivered to a place in deliver agent step 1422 (Figure 14B). It is reiterated that, in the illustrative example of Figures 15A-15E, agent 150A is not delivered to any place within engine 132A. However, for completeness, the delivery of agent 150A to a place within engine 132A is discussed. Agent 150A (Figure 15A) is delivered to a place within engine 132A when that place is defined by ticket 1306 (Figure 13A) as the trip destination.
Delivering an agent to a place involves (i) selecting a particular place to deliver the agent, (ii) making the agent an occupant of that place, and (iii) causing operation "go" to succeed for that agent. Deliver agent step 1422 is represented by logic flow diagram 1422 (Figure 14E).
Processing according to logic flow diagram 1422 (Figure 14E) begins with for all current places step 1422A. For all current places step 1422A and next step 1422C define a processing loop in which each of the places currently processed by the current engine, i.e., engine 132A (Figure 15A), is processed according to steps 1422B, 1422D, and 1422E (Figure 14E). In the context of the following discussion of logic flow diagram 1422, the "subject place" is the place which is processed in a particular iteration of the loop defined by for all current places step 1422A and next step 1422C. For each iteration of the loop, i.e., for each current place of engine 132A (Figure 15A), processing transfers from for all current places step 1422A to a place satisfies ticket test step 1422B.
In place satisfies ticket test step 1422B, the current engine determines whether the subject place satisfies ticket 1306.
If the subject place does not satisfy ticket 1306, either (i) processing transfers from place satisfies ticket test step 1422B (Figure 14E) through next step 1422C to for all current places step 1422A in which another iteration of the loop is begun, or (ii) processing transfers to terminal step 1422I, which is described below. Conversely, if the subject place satisfies ticket 1306, processing transfers from place satisfies ticket test step 1422B to entering step 1422D.
As described below in greater detail, a place grants ingress to a process, e.g. an agent, by successfully performing operation "entering" and denies ingress to a process by throwing an exception during performance of ~1~3~'~2 operation "entering". Performance of operation "entering"
is discussed in greater detail below.
Processing transfers from entering step 1422D (Figure 14E) to exception test step 1422E in which the current engine determines whether performance of operation "entering" by the subject place threw and exception. If performance of operation "entering" by the subject place threw an exception, thereby denying agent 150A ingress to the subject place, processing transfers from entering step 1422E through next step 1422C to for all current places step 1422A in which another iteration of the loop defined by for all current place step 1422A and next step 1422C is begun or processing transfers to terminal step 1422J, which is described below.
If, on the other hand, operation "entering" is performed successfully by the subject place, the subject place thereby granting to agent 150A (Figure 15A) ingress to the subject place, processing transfers from exception test step 1422E (Figure 14E) to step 1422F. In step 1422F, agent 150A (Figure 150A) is made an occupant of the subject place. Furthermore, a property of agent 150A is set to indicate that agent 150A is an occupant of the subject place.
Processing transfers from step 1422F to step 14226 in which operation "go" as performed by agent 150A succeeds.
As discussed above, agent 150A is not delivered to a place in engine 132A in the example of Figures 15A-15E.
Therefore, operation "go" for agent 150A does not succeed until agent 150A (Figure 15A) is delivered to place 220B
in engine 132B (Figure 15E) as discussed below.
Processing transfers from step 14226 (Figure 14E) to terminal step 1422H in which processing according to logic flow diagram 1422, and therefore deliver agent step 1422 (Figure 14B), completes successfully.
If all places which are current in engine 132A
(Figure 15) are processed according to the loop defined by _87_ ~~~3~~2 for all current places step 1422A (Figure 14E) and next step 1422C and terminal step 1422H is not reached, i.e., if no place that is current in engine 132A (Figure 15A) satisfies ticket 1306 and grants agent 150A occupancy via successful performance of operation "entering", processing transfers from for all current place step 1422A (Figure 14E) to terminal step 1422I. In terminal step 1422I, an exception of class "Destination Unavailable" is thrown and logic flow diagram 1422, and therefore deliver agent step 1422 (Figure 14B), completes. Thus, deliver agent step 1422 delivers agent 150A (Figure 15A) to a place identified by ticket 1306 (Figure 13A) as the destination of a trip or throws an exception of class "Destination Unavailable".
As discussed above, in the example of Figures 15A-15E, engine 1322 is the transfer destination. Agent 150A
is therefore encoded to form encoded agent 150A-E (Figure 15B) in step 1432 (Figure 14B) which is then transferred to engine 1322 (Figure 15C) in transfer out step 1434 (Figure 14B).
Operation "Go" from the Perspective of an Intermediate Engine As described above, transfer of encoded agent 150A-E
(Figure 15B) from communication infrastructure 132A-CI, across communications link 102AZ, to communication infrastructure 1322-CI (Figure 15C) is initiated in transfer out step 1434 (Figure 14B). Appended to encoded agent 150A-E (Figure 21) is destination 150A-E-D, which includes properties "name", "address" and "data" which are telename 2102, teleaddress 2104 and octet string 2106, respectively. Telename 2102 and teleaddress 2104 define the transfer destination of encoded agent 150A-E. The transfer destination as defined by destination 150A-E-D
(Figure 21) can be either engine 1322 (Figure 15B) or engine 132B. In one embodiment of the present invention, _88_ ~~~3~7?
destination 150A-E-D (Figure 21), which is formed in form destination step 1430 (Figure 14B), defines engine 1322 (Figure 15B) as the transfer destination of encoded agent 150A-E. In that case, communications infrastructure 132A-CI transfers encoded agent 150A-E directly to communications infrastructure 1322-CI of engine 1322. In another embodiment of the present invention, destination 150A-E-D (Figure 21) defines engine 132B (Figure 15B) as the transfer destination of encoded agent 150A-E. In the latter case, communications infrastructure 132A-CI
contains information and logic which causes communications infrastructure 132A-CI to transfer encoded agent 150A-E, which is destined for engine 132B according to destination 150A-E-D (Figure 21), to communications infrastructure 1322-CI (Figure 15B) or engine 1322.
In either embodiment described above, when engine 1322 (Figure 15C) receives encoded agent 150A-E, engine 1322 performs a system operation determined in route agent step 1414 "transferIn" which is shown as logic flow diagram 1400-I (Figure 14F). Engine 1322 (Figure 15C) begins performance of system operation "transferIn" by producing a way object which defines a transfer to destination 150A-E-D (Figure 21) in step 1402-I (Figure 14F). Engine 1322 produces a way object in step 1402-I
(Figure 14F) by consulting a finder within engine 1322 (Figure 15C) as described above with respect to finder 2050 (Figure 20B).
Processing transfers from step 1402-I (Figure 14F) to a fourth way out is here test step 1404-I in which engine 1322 (Figure 15C) determines whether the way object produced in step 1402-I (Figure 14F) defines a transfer to engine 1322 (Figure 15C). If the produced way object does not specify engine 1322 (Figure 15C) as the transfer destination, i.e., if destination 150A-E-D (Figure 21) defines engine 132B (Figure 15B) as the transfer destination, processing transfers from fourth way out is -89_ here test step 1404-I (Figure l4Fs) to a second transfer out step 1406-I. In second transfer out step 1406-I, engine 1322 (Figure 15C) initiates transfer of encoded agent 150A-E according to destination 150A-E-D (Figure 21) in a manner directly analogous to that described above with respect to transfer out step 1434 (Figure 14B).
Processing transfers from second transfer out step 1406-I to step 1408-I. In step 1408-I, engine 1322 (Figure 15C) adds encoded agent 150A-E to a list of pending transfers as described above with respect to step 1438 (Figure 14B). Processing transfers from step 1408-I
(Figure 14F) to terminal step 1410-I in which system operation "transferIn" completes successfully.
Thus, if the way object produced in step 1402-I
defines a transfer to engine 132B (Figure 15C), encoded agent 150A-E is transferred according to the produced way object. If, on the other hand, the produced way object defines a transfer to engine 1322, processing transfers from fourth way out is here test step 1404-I (Figure 14F) to step 1412-I. In step 1412-I, engine 1322 (Figure 15C) extracts from encoded agent 150A-E ticket 1306 (Figures 13A and 18A) and forms a copy of ticket 1306. Processing transfers from step 1412-I to step 1414-I in which property "way" of the ticket copy is cleared, i.e., set to a nil. Processing transfers from step 1414-I to a second route agent step 1416-I in which engine 1322 produces a way object defining a transfer of encoded agent 150A-E. The process performed in step 1416-1 is identical to the process represented by logic flow diagram 1414 (Figures 14C and 14D). In the context of second route agent step 1416-I, ticket 1306 used in the above discussion of logic flow diagram 1414 is replaced by the ticket copy made in step 1412-I, and the logic flow diagram is processed as described above using the ticket copy. Upon completion of step 1416-I, processing transfers to a fifth way out is here test step 1418-I.

143~"~2 .
In fifth way out is here test step 1418-I, engine 1322 (Figure 15C) determines whether the way object produced in step 1416-I (Figure 14F) defines a transfer to engine 1322 (Figure 15C). If the produced way object defines a transfer to engine 1322, processing transfers from fifth way out is here test step 1418-I (Figure 14F) to a decode agent step 1428-I which is described below in greater detail. Conversely, if the produced way object does not define a transfer to engine 1322 (Figure 15C), processing transfers from fifth way out is here test step 1418-I (Figure 14F) to a second form destination step 1420-I. In the example of Figures 15A-15F, ticket 1306 (Figure 13A), which is extracted from encoded agent 150A-E
and copied, defines a trip to place 220B of engine 132B
(Figure 15C). Therefore, processing transfers to second form destination step 1420-I.
In form destination step 1420-I, engine 1322 (Figure 15C) forms, i.e., supersedes, destination 150A-E-D (Figure 21) in a manner that is directly analogous to that described above with respect to form destination step 1430 (Figure 14B). Processing transfers from second form destination step 1420-I (Figure 14F) to a third transfer out step 1422-I. In third transfer out step 1422-I, encoded agent 150A-E (Figure 15C) is transferred according to destination 150A-E-D (Figure 21) as described above with respect to second transfer out step 1406-I (Figure 14F) .
Processing transfers from third transfer out step 1422-I to step 1424-I. In step 1424-I, engine 1322 (Figure 15C) adds encoded agent 150A-E to a list of pending transfers as described above with respect to step 1438 (Figure 14B). Processing transfers from step 1424-I
(Figure 14F) to terminal step 1426-I in which system operation "transferIn" completes successfully. In the example of Figures 15A-15F, encoded agent 150A-E is transferred from engine 1322 to engine 132B (Figure 15D) in a manner that is directly analogous to the transfer of encoded agent 150A-E from engine 132A to engine 1322 as described above.
Operation "Go" from the Perspective of the Destination Enaine When engine 1328 receives encoded agent 150A-E in communication infrastructure 132A-CI (Figure 15D), engine 1328 performs system operation "transferIn" in a manner similar to that described above with respect to engine 1322. Engine 1328 produces a way object defining a transfer according to destination 150A-E-D (Figure 21) in step 1402-I (Figure 14F) and determines whether the produced way object defines a transfer to engine 1328 (Figure 15D) in fourth way out is here test step 1404-I
(Figure 14F). Since engine 1328 (Figure 15D) processes place 2208 which is the destination of the trip defined by ticket 1306 (Figure 13A), the produced way object defines a transfer to engine 1328 (Figure 15D). Therefore, processing by engine 1328 transfers from fourth way out is here test step 1404-I (Figure 14F) to step 1412-I.
In steps 1412-I, 1414-I and 1416-I, engine 1328 extracts and copies ticket 1306 (Figure 13A) from encoded agent 150A-E, clears property "way" of the ticket copy, and routes encoded agent 150A-E according to logic flow diagram 1414 (Figures 14C and 14D) except now the ticket copy is used for ticket 1306 in the above description. In fifth Way out is here test step 1418-I (Figure 14F), engine 1328 (Figure 15D) determines whether the way object produced in second route agent step 1416-I (Figure 14F) defines a transfer to engine 1328 (Figure 15D). Since engine 1328 (Figure 15D) processes place 2208, the produced way object defines a transfer to engine 1328 (Figure 15D). Therefore, processing by engine 1328 transfers from fifth way out is here test step 1418-I
(Figure 14F) to decode agent step 1428-I.

In decode agent step ~~~~I~ "engine 132B decodes agent 150A from encoded agent 150A-E (Figure 15D) in communication infrastructure 132B-CI by reverse application of the encoding rules described in Appendix B
and stores agent 150A in data portion 132B-D (Figure 15E).
As network 1500 can be heterogeneous, the form of agent 150A within engine 132A (Figure 15A) may be inconvenient for representing agent 150A within engine 132B (Figure 15E). As long as the standardized form of encoded agent 150A-E is used to transfer agent 150A
between engines 132A and 132B, engine 132B can represent agent 150A in whatever form is most convenient for engine 132B when not transferring agent 150A.
Processing by engine 132B transfers from decode agent step 1428-I (Figure 14F) to deliver agent step 1430-I. In deliver agent step 1430-I, agent 150A is delivered according to logic flow diagram 1422 (Figure 14E) as described above with respect to deliver agent step 1422 (Figure 14B). In the context of deliver agent step 1430-I, the ticket is now the actual ticket 1306 (Figure 13A) and not the postulated ticket. In deliver agent step 1430-I (Figure 14F), agent 150A (Figure 15E) is granted occupancy in place 220B and operation "go" as performed by agent 150A completes successfully.
Processing transfers from deliver agent step 1430-I
(Figure 14F) to terminal step 1426-I in which processing according to logic flow diagram 1400-I (Figure 14F), i.e., performance of system operation "transferIn", by engine 132B completes successfully.
Any exception thrown in the course of performing the steps of logic flow diagram 1400-I (Figure 14F) causes performance of the steps of logic flow diagram 1450-I
(Figure 14G). Agent 150A is decoded from encoded agent 150A-E (Figure 15D) in step 1452-I (Figure 14F). Since agent 150A arrived in engine 132B (Figure 15D), agent 150A
is not retained in a list of pending transfers as WO 95102219 PCTIUS94107397 _ ~1436r1 ~
described above with respect to engine 132A. Therefore, it is necessary to decode agent 150A to continue execution of agent 150A. Processing transfers from step 1452-I
(Figure 14G) to deliver agent step 1454-I in which agent 150A is delivered to purgatory as described above with respect to step 1426 (Figure 14B). Processing transfers from deliver agent step 1454-I (Figure 14G) to terminal step 1456-I in which processing according to logic-flow diagram 1450-I completes. Agent 150A, which occupies purgatory as a result of deliver agent step 1454-I, throws the exception which caused performance of logic flow diagram 1450-I, thereby causing operation "go" as performed by agent 150A to fail.
Thus, agent 150A (Figures 15A-15E), by performance of operation "go", directs movement, i.e., transportation, of agent 150A, and objects owned by agent 150A such as objects 140A and 140B, through network 1500. After performance of operation "go", interpretation of agent 150A continues with the instruction within agent 150A which immediately follows operation "go".
Figure 13B shows a portion of the execution state and of the internal state of agent 150A immediately following performance of operation "go" by agent 150A. At the top of stack 1304 is ticket stub 1308 which is the result produced by performance of operation "go". Place 220B is the property of agent 150A identifying the place occupied by agent 150A, thereby indicating that agent 150A occupies place 220B.
Ingress and Ectress~ Operations "Enterinct" and "Exiting"
As discussed above, place 220B grants or denies ingress to agent 150A by performance of operation "entering". The embodiment disclosed above with respect to logic flow diagram 1400 (Figures 14A and 14B) transports encoded agent 150A-E (Figures 15A-15E) to engine 132B before requesting that place 220B perform ~1~3~'~2 operation "entering". In another embodiment of the present invention, place 2208 is directed to perform operation "entering" before agent 150A is transported across communications link 102AZ, through communication infrastructure 1322-CI, across communications link 102ZB
to engine 132B. By performing operation "entering", place 2208 either grants or denies agent 150A permission to occupy place 2208. However, it is anticipated that the present invention will be used in wide-area networks with high latency.
The term "latency" is used herein as it is used in the art to denote the amount of time between the time information is first sent by the sender of the information and the time information is first received by the receiver of the information. In networks with high latency, even short messages can require substantial amounts of time to reach a destination. In such a case, requesting that place 2208 (Figure 15A) perform operation "entering" and awaiting a result involves sending a request across communications link 102AZ, through communication infrastructure 1322-CI, across communications link 102ZB
and through engine 1328 to place 2208 to perform operation "entering" and awaiting receipt of the result produced by operation "entering" by the reverse path. Doing so before transporting agent 150A to engine 1328 postpones the transportation of agent 150A to engine 1328 by an amount of time approximately equal to twice the latency between engines 132A and 1328. Therefore, to substantially improve the performance of the present invention, performance of operation "entering" by place 2208 is postponed until agent 150A is transported to engine 1328.
While operation "entering" is described in the context of agent 150A (Figure 15E) entering place 2208 as a consequence of operation "go", operation "entering" is also performed as a consequence of a process, either an agent or a place, entering place 2208 as a consequence of x.., the process's creation within place 220B. The creation of a process is discussed in greater detail in Appendix A.
Figure 22A gives the execution state of agent 150A
immediately prior to performance of operation "entering".
The execution state of a process, e.g., agent 150A, is described in greater detail below. Operation "entering"
is performed at the request of engine 132B, and the per-formance of "entering" is recorded in the execution- state of agent 150A (Figure 15E). The permit of either agent 150A or place 220B, preferably place 220B, may be debited by performance of operation "entering".
Frame 2200 is part of the execution state of agent 150A and records the dynamic state of operation "entering"
as performed by place 220B. Frame 2200 includes stack 2202 which is the current stack. Stack 2202 contains, from top to bottom, contact 2208, permit 2206 and ticket 2204.
Contact 2208 identifies agent 150A as the process attempting to enter place 220B. As discussed below in greater detail and in Appendix A, a contact has, among its properties, a property "subjectName" and a property "subject". Property "subject" of contact 2208, which is ordinarily a reference to agent 150A, is nil in operation "entering" so that place 220B is not given a reference to agent 150A before place 220B grants ingress to agent 150A.
Property "subjectName" of contact 2208 is a telename which identifies agent 150A as the agent requesting ingress to place 220B.
Permit 2206 is the proposed local permit of the agent attempting to "enter" place 220B. Permit 2206 is passed "byProtectedRef" and therefore cannot be altered by performance of operation "entering". The passing of arguments and results, and in particular passing by protected reference, i.e., "byProtectedRef", is described below in greater detail. An agent's local permit defines the capabilities of the agent while at a given place. For WO 95/02219 ~ ~ ~ ~ ~ ~ ~ PCTIUS94/07397 example, agent 150A can limit itself to a subset of the capabilities and allowances of its native permit while at place 220B by supplying to place 220B permit 2206 which defines the subset of capabilities and allowances.
' S Ticket 2204 is passed "byProtectedRef" and is equal to ticket 1306 used to travel to responding place 220B and informs responding place 220B that agent 150A is attempting to enter from another place. If ticket 2204 is a nil, a process is attempting to enter responding place 220B by the process's creation within responding place 220B.
The method by which a place determines whether to admit a process, i.e., the method implementing operation "entering", as defined by class "Place" always throws an exception that is a member of class "Occupancy Denied", thereby denying the process ingress to the responding place. However, users of the present invention define subclasses of class "Place" which grant entrance to processes under specific circumstances. Such is a practical necessity for agents to travel from one place to another.
As the implementation of operation "entering", as performed by a place, is defined by the particular method provided by or inherited by the class of which the place is an instance, each place can implement operation "entering" differently from other places in the network.
Logic flow diagram 2260 (Figure 22C) serves as an illustrative example of an implementation of operation "entering" and therefore of the steps taken by place 220B
in performance of operation "entering" in entering step 1448 (Figure 14B).
Contact 2208, permit 2206 and ticket 2204 are popped from stack 2202 in steps 2262, 2264 and 2266, respectively (Figure 22C). Processing then transfers to ticket test step 2272 in which ticket 2204 is compared to a nil. If ticket 2204 is a nil, agent 150A is being created locally _97_ ~143~'~2 and processing transfers to terminal step 2274 where operation "entering" completes successfully. However, it is apparent in view of the foregoing and the following that a place can be defined such that operation "entering"
fails under certain circumstances so as to prevent the creation of a process within the place.
If ticket 2204 is not nil, processing transfers from ticket test step 2272 to allowance test step 2276.-In allowance test step 2276, property "charges" of permit 2206 is produced by querying attribute "charges" of permit 2206 and property "charges" of permit 2206 is compared to 1000. If property "charges" of permit 2206 is greater than 1000, processing transfers from allowance test step 2276 to terminal step 2278 where an exception of class "Occupancy Denied" is thrown, denying the agent ingress into the responding place. If property "charges"
of permit 2206 is less than or equal to 1000, processing transfers from allowance test step 2276 to terminal step 2274 where operation "entering" terminates successfully, granting the agent ingress into the responding place. The exact number 1000 as a maximum permissible charges allowance in the example of Figure 22C
is chosen arbitrarily for purposes of illustration.
The method of logic flow diagram 2260 is illustrative of an appropriate implementation of operation "entering"
for places processed by an engine operating within a personal computer system with limited resources. As discussed above and in Appendix A, property "charges" of a permit is an allowance of processing. Thus, according to logic flow diagram 2260, agents, which indicate by a large charges allowance in their proposed local permits a need for large amounts of computation and resources, are denied ingress. Thus, the owner of a small personal computer system can restrict her personal computer system to processes created locally and to relatively inexpensive visits by agents created elsewhere.
_98_ WO 95/02219 ~ ~ PCT/US94107397 Immediately following performance of operation "entering" by place 220B (Figure 22B), stack 2202 is empty as operation "entering" produces no result.
As discussed above, a place notes the departure, i.e., the termination of occupancy, of a process by performance of operation "exiting". In the context of logic flow diagram 1400 (Figures 14A and 14B), place 220A
is performing operation "exiting" as a result of ari agent, i.e., agent 150A, leaving place 220A as a consequence of operation "go". Operation "exiting" is also performed by place 220A when a process, either an agent or a place, no longer occupies place 220A as a consequence of the process's destruction. The destruction of a process is described in greater detail in Appendix A. The interface of operation "exiting" as defined by class "Place" is shown by Figures 23A and 23B.
As operation "exiting" is performed at the request of engine 132A (Figure 15A), frame 2300 (Figure 23A), which records the dynamic state of operation "exiting", is neither part of the execution state of place 220A (Figure 15A) nor part of the execution state of agent 150A. In fact, at the time at which place 220A is performing operation "exiting", agent 150A no longer occupies place 220A. Engine 132A creates a new execution state as part of an "engine process". An engine process is a process created by an engine for the purpose of carrying out a performance of operation "exiting" or of operation "parting". Operation "parting" is discussed below and in Appendix A.
Figure 23A illustrates frame 2300 immediately prior to performance of operation "exiting". Frame 2300 includes stack 2302, which is the current stack.
Stack 2302 contains as the arguments of operation "exiting", from top~to bottom, contact 2308, permit 2306 and ticket 2304.

~143fi'~
Contact 2308 identifies agent 150A (Figure 15A) as the process exiting place 220A. As discussed below in greater detail and in Appendix A, a contact has, among its properties, a property "subjectName" and a property "subject". Property "subjectName" of contact 2308 (Figure 23A) identifies agent 150A (Figure 15A); property "subject" of contact 2308 (Figure 23A), which is ordinarily a reference to agent 150A (Figure 15A), is voided in operation "exiting" so that place 220A is not left with a reference to exiting agent 150A.
Permit 2306 (Figure 23A) is the current local permit of the process exiting responding place 220A (Figure 15A), i.e. of agent 150A. A process's local permit defines the capabilities of the process while occupying a given place.
Local permits are discussed in greater detail in Appendix A. Permit 2306 (Figure 23A) is passed "byProtectedRef" and therefore cannot be altered by place 220A (Figure 15A) in performing operation "exiting".
The local permit of the exiting process, i.e., permit 2306 (Figure 23A), is used by the engine interpreting responding place 220A (Figure 15A) to determine the type and amount of resources used by the exiting process during occupancy of place 220A so that the authority of the exiting process can be billed for use made of place 220A and resources contained therein. The type and amount of resources used are determined by comparison between permit 2306 (Figure 23A) and the permit consumed as an argument in operation "entering" when granting ingress to the exiting process. Operation "entering" is discussed in greater detail above and in Appendix A.
For example, place 220A (Figure 15A) in performing operation "entering", can store the proposed local permit, e.g., permit 2206 (Figure 22A), for later comparison with permit 2306 (Figure 23A), which is consumed in performance of operation "exiting". As described below and in Appendix A, a contact, e.g., contact 2308, has a property "subjectNotes". In one embodiment, place 220A stores, within property "subjectNotes" of contact 2208 (Figure 22A), permit 2206 in performing operation "entering". In performing operation "exiting", place 220A
(Figure 15A) determines the amount of resources consumed by agent 150A while occupying place 220A by comparing permit 2206 (Figure 22A), which is stored within property "subjectNotes" of contact 2208, with permit 2308 (Figure 23A), which is consumed in performance of operation "exiting".
Ticket 2304 (Figure 23A) is passed "byProtectedRef"
and is equal to ticket 1306 (Figure 13A) which is used by agent 150A (Figure 15A) to specify the destination of the trip and informs responding place 220A that agent 150A is leaving responding place 220A as a consequence of performing operation "go". If ticket 2304 (Figure 23A) is a nil, agent 150A (Figure 15A) is exiting responding place 220A as a consequence of the destruction of agent 150A.
Frame 2300 (Figure 23B) is shown immediately following performance of operation "exiting" by place 220A
(Figure 15A). Stack 2302 (Figure 23A) is empty as operation "exiting" produces no result. As frame 2300 (.Figure 23B) is not part of the execution state of either place 220A (Figure 15A) or agent 150A, but is instead part of the execution state of an engine process, any exception thrown by performance of operation "exiting" is not experienced by, and therefore has no effect upon, either place 220A or agent 150A. In performing operation "exiting", place 220A notes the departure of agent 150A
and objects 140A and 140B, which are owned by agent 150A.
Obiect Interchange The time required to transport an agent from one place to another is substantially reduced by limiting transportation of objects contained by the agent to only ~143~'~~
those objects that are not likely to have an equivalent object at the destination place. Limiting transportation of such objects is particularly important as agent 150A, in traveling from engine 132A to engine 1322, includes class objects representing the classes of which agent 150A
and objects owned by agent 150A are members. As class objects are typically quite large and as the classes of which an agent and the objects owned by the agent are members are typically quite numerous, transporting all such class objects is impractical. Thus, it is preferred that, in transporting agent 150A from engine 132A to engine 132B (Figure 15A), only class objects representing classes, which are not represented by class objects in engine 132B, are transported with agent 150A to engine 132B.
Objects, that are likely to have equivalent objects at most places within network 1500 (Figures 15A-15E), are members of mix-in class "Interchanged" and therefore include a property "digest". Members of mix-in class "Interchanged" are called "interchanged objects". The object which is property "digest" of an interchanged object is alternatively called "the digest" of the interchanged object.
An interchanged object is one that an engine, e.g.
engine 132A or 132B (Figure 15A), deems equivalent to any other instance of the interchanged object's class whose digest is equal to the digest of the interchanged object.
A digest is any object suited to the purpose of distinguishing a first interchanged object from all others. For example, a mathematical hash of a canonical binary representation of the interchanged object is suitable for many classes of interchanged objects. Class objects, i.e., objects of class "Class", define built-in classes as well as user-defined classes, as described more completely below and in Appendix A, and do not vary from WO 95/02219 s - ~ 1 ~ 3 6 7 ~ PCT/US94/07397 engine to engine within computer network 1500. Class objects are therefore interchanged objects.
The use of digests to reduce the amount of data transferred, and therefore to reduce the time required to transport an agent and the objects contained by that agent from one place to another, is illustrated by logic flow diagrams 2400 and 2450 (Figures 24A and 24B).
Logic flow diagram 2400 (Figure 24A) represents the process of transporting agent 150A (Figure 15A) and the objects contained by agent 150A from a source engine 132A
to a destination engine 132B when interchanged objects are considered. An agent "contains" all objects owned by the agent as well as every class of which the agent is a member and every class of which any object owned by the agent is a member. As described above, agent 150A and all objects contained by agent 150A are packaged into a shipping box, encoded and transported to engine 132B. In packaging agent 150A and objects contained therein, source engine 132A collects all objects contained by agent 150A
in step 2402 (Figure 24A). Processing transfers from step 2402 to for each object step 2404 which, with next step 1412, defines a loop wherein each object collected in step 2402 is considered. Processing transfers from for each object step 2404 to digest test step 2406 in which the class membership of each object contained by agent 150A is checked to determine whether each object has a digest.
As discussed above and in greater detail in Appendix A, members of mix-in class "Interchanged" can include a digest. Therefore, if an object has a digest, the object is an interchanged object. If an interchanged object does not have a digest, i.e., if attribute "digest" of the interchanged object is a nil, the interchanged object is not interchanged but is instead treated as a non-interchanged object.

WO 95/02219 , PCT/US94/07397 ~143~'~2 If an object has no digest, even if the object is an interchanged object, processing transfers from digest test step 2406 to step 2410 in which the object is encoded according to the encoding rules of Appendix B and included in the shipping box. Otherwise, if the object has a digest, processing transfers from digest test step 2406 to step 2408 in which the digest of the object and a citation of the class of which the object is an instance are encoded according to the encoding rules of Appendix B and included in a parts box which is in turn included in the shipping box.
Processing transfers from step 2408 or step 2410 to next step 2412 in which processing is transferred to for each object step 2404. Engine 132A repeats the process of steps 2406, 2408 and 2410 for each object contained by agent 150A in a loop defined by for each object step 2404 and next 2412. If all objects collected in step 2402 have been processed according to the loop of for each object step 2404 and next object step 2412, processing transfers from for each object step 2404 to a transfer out encoded agent step 2414.
In transfer out encoded agent step 2414, the shipping box is encoded and is transported by source engine 132A to destination engine 132B, as described above. Thus, the encoded agent which is transported to destination engine 132B includes all objects contained by the agent which have no digests and the digests of all objects contained by the agent which are interchangeable and which have digests. Transfer out encode agent step 2414 is represented in double boxes to indicate interaction between engines across the network. Such an exchange of binary data between engines is discussed above and in Appendix C.
Logic flow diagram 2450 (Figure 24B), illustrates the transfer of encoded agent 150A from the perspective of destination engine 132B. Destination engine 132B receives WO 95/02219 ~ ~ 4 3 6 ~~ ~ PCT~S94/07397 the encoded agent as binary data from source engine 132A
in transfer in encoded agent step 2452. As described above in the context of Figures 15A-15E, the encoded agent can pass through one or more intermediate engines, e.g., engine 1322, enroute from engine 132A to engine 1328.
Processing transfers from transfer in encoded agent step 2452 to step 2454 in which the encoded agent is decoded, thereby reconstructing the shipping box whose contents are the agent objects contained by the agent.
Each of the objects, which do not have digests and which are contained by the agent, and the digests of the interchanged objects contained by the agent which have digests, are processed by destination engine 1328 according to the loop formed by for each object step 2456 and next step 2462. Each iteration of the loop processes one of the objects decoded in step 2454.
Processing transfers from for each object step 2454 to digest test step 2464. In digest test step 2464, destination engine 1328 determines whether an object is a digest. If the object is not a digest, processing transfers from digest test step 2464 directly to next object step 2462. Conversely, if the object is a digest, processing transfers from digest test step 2464 to test step 2466 and the object is called "the subject digest".
In test step 2466, destination engine 1328 determines whether the subject digest is equivalent to a digest of an interchanged object which is present in engine 132B. The method by which such a determination is made is discussed in more detail below.
If an interchanged object having an equivalent digest is found in step 2466, processing transfers from test step 2466 to substitute step 2468. In substitute step 2468, the interchanged object is substituted for the subject digest by replacing all references within the objects decoded in step 2454 to the interchanged object represented by the subject digest with references to the found interchanged object. Processing transfers from substitute step 2468 to next object step 2462.
If, on the other hand, no interchanged object having an equivalent digest is found in test step 2466, processing transfers from test step 2466 to step 2470. In step 2470, the subject digest is added to a list of digests. A new list is created if no such list exists.
Processing transfers from step 2470 to next step 2462.
From next step 2462, processing transfers to for each object step 2456. If all objects decoded in step 2454 have been processed according to the loop of for each object step 2456 and next step 2462, processing transfers from for each object step 2456 to a list test step 2476, which is described below.
An engine searches for equivalent interchangeable objects in step 2466 as follows. Each engine contains one or more places which can exchange equivalent interchangeable objects. Each such place includes a "repository" of interchangeable objects which are present within the place. Dictionary 2490 (Figure 24D) is the repository of place 220B (Figure 15D). The keys of dictionary 2490 (Figure 24D) are classes, i.e., classes 2490K1, 2490K2, 2490K3 and 2490K4. Associated with each class is a dictionary; for example, associated with classes 2490K1-2490K4 are dictionaries 2490A, 2490B, 2490C
and 2490D, respectively.
Each of dictionaries 2490A-2490D have the same general organization; dictionary 2490A is illustrative.
The keys of dictionary 2490A are digests, i.e., digests 2490AK1, 2490AK2 and 2490AK3. Associated with each digest is an interchanged object whose digest is the associated digest; for example, associated with digests 2490AK1-2490AK3 are interchanged objects 2490AV1, 2490AV2 and 2490AV3, respectively.
An interchanged object is retrieved from dictionary 2490 in two steps. As discussed above, both a digest and - ~143~'~~
a citation of a class are used within a parts box to represent an interchanged object. In the first step, a citation is used to retrieve the dictionary of dictionaries 2490A-2490D which is associated with the class referenced by the citation. For example, if the citation references class 2490K1, dictionary 2490A is retrieved.
In the second step, the retrieved dictionary is searched for an association between a digest equivalent to the digest which is included in the parts box and an interchanged object. If an association is found, the interchanged object is retrieved. If no association is found, the place contains no interchanged object which is equivalent to the interchanged object represented by the citation and digest in the parts box. Thus, a repository such as dictionary 2490 (Figure 24D) is used to determine whether a place contains an equivalent interchanged object and to retrieve such an object.
As described above, processing transfers from for each object step 2456 to list test step 2476. In list test step 2476, the destination engine, e.g. engine 132B, determines whether a list of digests was created in step 2470 and, if so, whether the list is empty. If the list of digests does not exist or is empty, processing transfers from list test step 2476 to activate traveling agent step 2478 in which engine 132B activates traveling agent 150A by scheduling agent 150A for execution.
Processing transfers from activate traveling agent step 2478 to terminal step 2479 in which the transfer of agent 150A, as carried out by engine 132B, completes successfully.
If, on the other hand, a list was created in step 2470 and the list contains at least one digest, an equivalent object for at least one interchanged object is not found within destination engine 132B and processing transfers from list test step 2476 to hold traveling agent ~~43~'~2 step 2471. In hold traveling agent step 2471, traveling agent 150A is held in a suspended state in a holding queue within engine 1328. Processing transfers from hold traveling agent step 2471 to step 2472.
In step 2472, the destination engine, e.g. engine 1328, creates an object retrieval agent, i.e., an "OR.A".
The ORA retrieves interchanged objects from the source engine, e.g., engine 132A according to logic flow diagram 2480 (Figure 24C). Processing transfers from step 2472 to step 2474 in which the ORA is directed to perform operation "live", thereby initiating interpretation of the ORA. As discussed below in greater detail, the central activity of the ORA is shown by logic flow diagram 2480 (Figure 24C). Processing transfers from step 2474 to terminal step 2479 in which the transfer of agent 150A, as carried out by engine 1328, completes successfully. It should be noted that traveling agent 150A is not activated and is in the form of a shipping box. Agent 150A is reconstituted from the shipping box and reactivated by the ORA as described below.
Logic flow diagram 2480 (Figure 24C) illustrates the central procedure of the ORA. In step 2482, the ORA
travels to the source engine, e.g., engine 132A, by performance of operation "go". Processing transfers from 2482 to step 2484 in which the ORA collects from an interchanged object repository, such as dictionary 2490 (Figure 24D?, within engine 132A a copy of each object having a digest that is equivalent to a digest contained in the list of digests. Processing transfers from step 2484 to step 2485.
In step 2485, the ORA travels back to the destination engine, e.g., engine 1328, by performance of operation "go". The ORA carries to the destination engine copies of those objects whose digests are contained in the list of digests, namely, those objects for which there are no equivalent objects within the destination engine.

WO 95!02219 PCTIUS94I07397 ~143~~2 Processing transfers from step 2485 to step 2486. In step 2486, the ORA substitutes the objects collected for the digests within agent 150A as described above.
Processing transfers from step 2486 to activate traveling agent step 2487, in which agent 150A is reconstituted from the shipping box and activated as described above. The reconstitution of an agent and the objects contained by the agent from a shipping box is described in greater detail in Appendix D. Processing transfers from step 2487 to terminal step 2488 in which the central procedure of the ORA is successfully completed.
Thus, efficiency is achieved by transporting across network communications media only those interchanged objects which are not equivalent to interchanged objects within the destination engine.
Operation "Send"
An agent is capable of traveling to several places simultaneously by performing operation "send". Figure 25 shows the state of network 2500 prior to performance of operation "send" by agent 150A. In the example of Figure 25, agent 150A in data portion 132A-D of engine 132A is configured to transport a clone of itself to place 220B in data portion 132B-D of engine 132B and a clone of itself to place 220C in data portion 132C-D of engine 132C
simultaneously. In other words, agent 150A is performing operation "send", supplying as the argument to operation "send" a list of two tickets specifying places 220B and 220C as destination places for clones of agent 150A. A
clone of agent 150A is a copy of agent 150A which is made by replicating agent 150A, including the execution state of agent 150A, to form the copy and which is made "active"
by assigning a new name and identifier to the copy and scheduling the copy for execution by an engine. Clones are described in greater detail in Appendix A.

The interface of operation "send" is illustrated by Figures 26A, 26B and 26C. Figure 26A illustrates frame 2602, which records the dynamic state of operation "send"
as performed by agent 150A and which is a part of the execution state of agent 150A, immediately prior to performance of operation "send". The execution state of a process, including an agent, is described in greater detail below and in Appendices A and B. -Included in frame 2602 is stack 2604. Stack 2604 is the stack from which arguments are popped and to which a result is pushed during performance of operation "send".
Stack 2604 is therefore the "current stack" in the context of operation "send". The "current stack" is defined and explained more completely below and in Appendices A and B.
At the top of stack 2604, as indicated by letter "T", is list 2606 whose items are tickets 2608 and 2610.
Tickets 2608 and 2610 of list 2606 each define a trip to be taken by a corresponding clone of agent 150A.
Objects 2618 and 2620 are property "travelNotes" of tickets 2608 and 2610, respectively. Objects 2618 and 2620 provide a mechanism by which respective clones created by performance of operation "send" can be distinguished. For example, in forming tickets 2608 and 2610, agent 150A can store within objects 2618 and 2620 strings whose respective texts are "Able" and "Baker". In such a case, the clone of agent 150A corresponding to ticket 2608 retrieves the string "Able" from property "travelNotes" of ticket 2608. Similarly, the clone of agent 150A corresponding to ticket 2610 retrieves the string "Baker" from property "travelNotes" of ticket 2610.
The particular implementation by which the respective clones are distinguished is ultimately up to the user of the present invention. The following is another example.
Stored within object 2618 is an integer whose value is one, thereby representing the position of corresponding ticket 2608 within list 2606. Similarly, stored within 143~7~
object 2620 is an integer whose value is two, thereby representing the position of corresponding ticket 2610 within list 2606.
Permits 2622 and 2624 are properties "destinationPermit" of tickets 2608 and 2610, respectively. Integers 2626 and 2628 are properties "charges" of permits 2622 and 2624, respectively.
Permit 2612 is property "permit" of agent 150A and therefore represents a part of the internal state of agent 150A. As described above and in Appendix A, permit 2612 limits the execution of agent 150A.
Boolean 2614 is property "canSend" of permit 2612. If the value of Boolean 2614 is "false", operation "send" fails and throws an exception of class "Permit Violated".
Property "charges" of permit 2612 is integer 2616. As discussed in greater detail in Appendix A, integer 2616 represents the processing allowance of agent 150A.
In the example of Figure 25, agent 150A, whose current location is place 220A, is configured to transport respective clones of agent 150A to places 220B and 220C.
As agent 150A sends a clone of itself to each destination place defined by a ticket in list 2606 (Figure 26A), performance of operation "send" eventually creates as many clones of agent 150A as there are tickets in list 2606.
Figures 27A and 27B illustrate a simple embodiment of operation "send" wherein engine 132A (Figure 27A) creates within data portion 132A-D a number of clones of agent 150A equal to the number of tickets in list 2606 (Figure 26A). In this example, list 2606 (Figure 26A) contains two tickets, i.e., tickets 2608 and 2610.
Therefore, engine 132A creates two clones, agent 150A-1 and agent 150A-2, of agent 150A (Figure 27B). A more efficient embodiment, in which cloning of the responding agent is deferred, is described below.
In forming the clones of agent 150A, permits 2622 and 2624 are made property "permit" of the clones created. In WO 95/02219 PCTIUS94107397 _ ~143G72 other words, permit 2622 is property "permit" of agent 150A-1 and permit 2624 is property "permit" of agent 150A-2. In creating agents 150A-1 and 150A-2, the allowances of permits 2622 and 2624 are subtracted from the corresponding allowances of permit 2612 of agent 150A, i.e., the agent that is being cloned. In other words, the value of integers 2626 and 2628 are subtracted from integer 2616, which is the charges allowance of agent 150A, i.e., property "charges" of the permit of agent 150A, upon creation of agents 150A-1 and 150A-2. If integer 2616 is not greater than or equal to the sum of integers 2626 and 2628, operation "send" fails and throws an exception of class "Permit Violated" since the charges allowance of agent 150A is less than the sum of the charges allowances of the clones created from agent 150A.
Thus, the permits which are part of the tickets supplied as arguments to operation "send" form the native permits of the clones created in performance of operation "send". As tickets 2608 and 2610 define trips to be taken by respective clones of agent 150A, permits 2622 and 2624 are property "permit" of tickets 2608 and 2610, respective, and are therefore also the local permits of respective clones of responding agent 150A.
In a second embodiment of the present invention, operation "send" consumes, as an argument in addition to list 2606, an integer (not shown). In creating clones of responding agent 150A, respective native permits are formed as described above with the exception that property "charges" of each native permit is initially equal to the integer consumed.
A third embodiment of the present invention is the same as the second embodiment described above with the exception that a list of integers is consumed in lieu of a single integer. Additionally, property "charges" of each native permit, which is derived from property "permit" of a ticket within list 2606 (Figure 26A), is initially equal ~I~3~'~2 to the integer at a position within the consumed list of integers that is equal to the position within list 2606 of the ticket from which the native permit is derived. For example, property "charges" of the native permit of the clone corresponding to ticket 2608 at position one within list 2606 is initially equal to the integer at position one in the consumed list of integers (not shown).
Similarly, property "charges" of the native permit-of the clone corresponding to ticket 2610 at position two within list 2606 is initially equal to the integer at position two in the consumed list of integers.
In a fourth embodiment of the present invention, operation "send" consumes, as an argument in addition to list 2606, a permit (not shown). In creating clones of responding agent 150A, each respective native permit is a copy of the consumed permit.
A fifth embodiment of the present invention is the same as the fourth embodiment described above with the exception that a list of permits is consumed in lieu of a single permit. Additionally, each native permit of a respective clone corresponding to a ticket within list 2606 (Figure 26A) is initially equal to the permit at a position within the consumed list of permits that is equal to the position within list 2606 of the ticket to which the clone corresponds. For example, the native permit of the clone corresponding to ticket 2608 at position one within list 2606 is initially equal to the permit at position one in the consumed list of permits (not shown).
Similarly, the native permit of the clone corresponding to ticket 2610 at position two within list 2606 is initially equal to the permit at position two in the consumed list of permits.
In each of the described embodiments, property "charges" of permit 2612, which is the permit of responding agent 150A, is reduced by the total of respective properties "charges" of the native permits of ~1~3~'~2 the clones of agent 150A created in performance of operation "send".
Each clone of agent 150A is transported to a place identified by the corresponding ticket as a destination place. For example, ticket 2608 identifies place 220B as the destination for agent 150A-1, and ticket 2610 identifies place 220C as the destination for agent 150A-2.
Each clone travels to its respective destination place through the cooperation of the source engine 132A with a corresponding destination engine, e.g., either engine 132B
or engine 132C. The transportation of an agent clone, e.g. either agent 150A-1 or agent 150A-2, from one place to another in performing operation "send" is as described above in detail in conjunction with operation "go".
If, in performance of operation "send", agent 150A
successfully creates agents 150A-1 and 150A-2, performance of operation "send" by agent 150A succeeds even if the transportation of any or all of the agent clones fails.
If a trip of an agent clone fails, the trip exception is thrown by the agent clone, not the agent which originally performed operation "send".
A portion of the execution state of agent 150A
immediately following performance of operation "send" by agent 150A is shown in Figure 26B. Stack 2604, which is the current stack as described above, contains a nil object 2630. A nil object is produced as a result for the original agent, thereby distinguishing the original agent from agent clones created in performance of operation "send".
Figure 26C shows a portion of the execution state of agent 150A-1 after performance of operation "send" by agent 150A. The execution state of agent 150A-2 after performance of operation "send" by agent 150A is directly analogous to the execution state of agent 150A-1 as described immediately below.

~~~3~72 Permit 2622 is property "permit" of agent 150A-1.
Stack 2604-1 is a copy of stack 2604 produced in the creation of agent 150A-1. Stack 2604-1 is the current stack of agent 150A-1. At the top of stack 2604-1 is ticket stub 2632. Ticket stub 2632 is the result produced by operation "send" for an agent clone created by performance of operation "send". Ticket stub 2632 is derived from ticket 2608.
Ticket stub 2632 is a member of a class "Ticket Stub". Ticket stub 2632 has, among other properties, properties "way" and "travelNotes". Ticket 2608 is a member of a class "Ticket". Class "Ticket" is a subclass of class "Ticket Stub"; therefore, ticket 2608 has, among other properties, properties "way" and "travelNotes", definitions of which are inherited from superclass "Ticket Stub". Ticket stub 2632 is derived from ticket 2608 such that properties "way" and "travelNotes" of ticket stub 2632 are equal to properties "way" and "travelNotes", respectively, of ticket 2608. As discussed above, property "travelNotes" of ticket 2608 (Figure 26A) can be used to distinguish agent 150A-1 from agent 150A-2.
Similarly, the current stack (not shown) of agent 150A-2 contains a ticket stub (not shown) whose property "travel Notes" is object 2620.
Figure 28 shows computer network 2500 after operation "send" has been performed by agent 150A. Agent 150A
remains in data portion 132A-D of computer process engine 132A. Agent 150A-1 occupies place 220B in data portion 132B-D of engine 132B. Similarly, agent 150A-2 occupies place 220C executing in data portion 132C-D of engine 132C.
The access of operation "send" is "private"
permitting only the responder to request the operation.
As discussed in greater detail below and in Appendix A, each feature of the present invention has an access which specifies under what conditions the feature can be WO 95/022 ~; PCT/US94/07397 requested. As the access of operation "send" is private, only agent 150A can initiate performance of operation "send" by agent 150A. Even engine 132A cannot initiate performance of operation "send" by agent 150A.
Deferred Cloninct Substantial savings in the amount of time required to perform operation "send" and in space occupied by clones of an agent created during performance of operation "send"
are realized by "deferred cloning". Deferred cloning occurs when the engine carrying out performance of operation "send" determines that two or more clones of the same agent are to be transferred to a single destination engine. In such a case, a single clone is transferred to the destination engine and additional clones are derived from the single clone at the destination engine.
Figures 29A-29D are illustrative.
Agent 150A (Figure 29A) occupies place 220A in engine 132A. Engine 132A is a computer process executing within computer system 110A (not shown). Agent 150A is performing operation "send", supplying as the argument to the operation a list of three tickets (not shown) specifying places 220B, 220C and 220D as destination places for clones of agent 150A. Engine 132A is in communication with engine 132E across communications link 102AE. Communications link 102AE is as described above with respect to communications link 102AZ (Figure 15A) in conjunction with operation "go". In addition to engine 132A, engine 132E is in communication with engines 132B
and 132F across communications link 102EBF. Additionally, engine 132F is in communication with engines 132C and 132D
across communications link 102FCD. Places 220B, 220C and 220D are processes interpreted by engines 132B, 132C and 132D, respectively. Thus, for clones of agent 150A to reach places 220B, 220C and 220D, each clone must first be transferred to engine 132E.

21430'2 Created within engine 132A is a single clone, i.e., agent 150A-1, of agent 150A. The execution state of agent 150A-1 includes a send frame 2902 as agent 150A-1 is performing operation "send". The execution state of an agent is discussed in greater detail below and in Appendix A. Send frames are discussed in greater detail in Appendix B. Send frame 2902 (Figure 30A) includes list of tickets 2904 which includes tickets 2904B, 2904C and 2904D which in turn specify places 220B, 220C and 220D, respectively, as destination places. Agent 150A-1 (Figure 29A), the single clone of agent 150A, therefore represents three separate agent clones which are to travel to places 220B, 220C and 220D, respectively.
Encoded agent 150A-1-E (Figure 30B) is the result of encoding agent 150A-1 (Figure 29A) according to the encoding rules of Appendix B. Included in encoded agent 150A-1-E (Figure 30B) are destinations 150A-1-E-D1, 150A-1-E-D2, and 150A-1-E-D3 which define respective transfer destinations of the three clones of agent 150A
(Figure 29A). As all three clones of agent 150A are transferred to engine 132E, destinations 150A-1-E-D1, 150A-1-E-D2, and 150A-1-E-D3 (Figure 30B) all define engine 132E as the transfer destination of a respective clone of agent 150A (Figure 29A). Engine 132A determines that destinations 150A-1-E-D1, 150A-1-E-D2, and 150A-1-E-D3 (Figure 30B) each define engine 132E as a transfer destination and transfers a single clone of agent 150A (Figure 29A), i.e., encoded agent 150A-1-E (Figure 30B) to engine 132E (Figure 29B). Thus, rather than transporting three separate clones from engine 132A to engine 132E, a single clone is transported to engine 132E
reducing proportionally the amount of space within engine 132A occupied by clones of agent 150A and the amount of time required to transport clones to engine 132E.
Agent 150A-1 is encoded and transferred to engine 132E (Figure 29B) as discussed above in conjunction with ~1436'~~
operation "go". Engine 132E retrieves list 2904 of send frame 2902 of agent 150A-1. As the structure of an encoded agent is standardized as described in Appendix B, an engine, e.g., engine 132E, can retrieve list 2904 without decoding encoded agent 150A-1. Engine 132E
determines that one clone corresponding to ticket 2904B of send frame 2902 is to be transported to engine 132B and that two clones corresponding to tickets 2904C and-2904D
are to be transported through engine 132F to engines 132C
and 132D, respectively. Therefore, a second clone, agent 150A-2 (Figure 29C), is created from agent 150A-1 by engine 132E. As agent 150A-1 is encoded, agent 150A-2, which is a copy of agent 150A-1, is encoded as well.
Agent 150A-2 includes send frame 2902-2, and send frame 2902-2 in turn includes list of tickets 2904-2.
List of tickets 2904-2 includes a single ticket 2904B
which is removed from list of tickets 2904 of send frame 2902 of agent 150A-1.
Agent 150A-2 is transferred to engine 132B, and agent 150A-1 is transferred to engine 132F (Figure 29D).
As ticket 2904B specifies place 220B as the destination of a trip and ticket 2904B is the only ticket in send frame 2902-2, place 220B is the destination of agent 150A-2 and no further clones of agent 150A are created from agent 150A-2. Agent 150A-2 is decoded by engine 132B. Agent 150A-2 is granted occupancy to place 220B by performance of operation "entering" by place 220B
as described above, and performance of operation "send"
for agent 150A-2 is completed. Send frame 2902B is therefore removed from the execution state of agent 150A-2 as discussed below in the context of the execution model.
Agent 150A-1 is transferred to engine 132F. Included in send frame 2902, which is part of the execution state of agent 150A-1, is list of tickets 2904 which in turn includes tickets 2904C and 2904D. Thus, agent 150A-1 represents two clones of agent 150A which are traveling to ~1~~~'~2 places 220C and 220D, respectively. Therefore, by transferring a single clone, i.e. agent 150A-1, from engine 132E to engine 132F, substantial savings are realized in storage space in engine 132E and in time ' 5 required to transport data between engine 132E and engine 132F as described above.
In the manner described above, in engine 132F (i) a third clone of agent 150A, i.e., agent 150A-3, is formed from agent 150A-1, and (ii) a send frame (not shown) is included in agent 150A-3 which includes ticket 2904C. As agent 150A-1, from which agent 150A-3 is copied, is encoded, agent 150A-3 is encoded as well. Ticket 2904C is removed from send frame 2902 of agent 150A-1.
Agent 150A-3 is transferred to engine 132C, and agent 150A-1 is transferred to engine 132D (Figure 29E).
Engine 132C includes place 220C which is the trip destination of agent 150A-3 as defined by ticket 2904C, and engine 132D includes place 220D which is the trip destination of agent 150A-1 as defined by ticket 2904D.
In the manner described above, (i) engine 132C decodes agent 150A-3 and agent 150A-3 is granted occupancy of place 220C by performance of operation "entering" and (ii) engine 132D decodes agent 150A-1 and agent 150A-1 is granted occupancy of place 220D by performance of operation "entering". Since send frame 2902 (not shown) contains only a single ticket which specifies place 220D
as the destination, operation "send" completes for agent 150A-1. Similarly, as agent 150A-3 includes a send frame specifying a single place, namely, place 220C which is the place occupied by agent 150A-3, as the destination of a trip, operation "send" completes for agent 150A-3 as well.
Each of agents 150A-1, 150A-2 and 150A-3 completes performance of operation "send" having an execution state which includes a send frame which includes a single ticket which defines the trip taken by the respective agent ~I43fi'~~
clone. The respective single ticket is used as described above to derive a ticket stub which is produced as the result of the respective agent's performance of operation "send".
Thus, substantial amounts of computer storage space and time are saved in performance of operation "send" by deferring cloning as long as possible in the transportation of the several clones to respective destination places.
The structure of send frame 2902 during transport of agent 150A-1 is shown in Figure 30. Class "Send Frame"
defines property "tickets". Property "tickets" of send frame 2902 is list 2904 whose items are tickets 2904B, 2904C and 2904D. Tickets 2904B, 2904C and 2904D define respective trips to be taken by respective clones of responding agent 150A.
As discussed above, performance of operation "send"
by agent 150A creates one or more agent clones, each agent clone representing one or more clones of agent 150A. If forming each agent clone, e.g. agent 150A-1 (Figure 29A), a copy of send frame 2902 is included in the agent clone, Property "tickets" of the send frame copy is modified so as to include only those tickets corresponding to the clones represented by the agent clone. For example, agent 150A-2 (Figure 29C) includes send frame 2902-2 which in turn includes ticket 2904B and agent 150A-1 includes send frame 2902 which in turn includes tickets 2904C and 2904D. Therefore, agent 150A-2 represents a single clone of agent 150A which is traveling to the place specified by ticket 2904B, namely, place 220B. Similarly, agent 150A-1 represents two clones of agent 150A which are traveling to places specified by tickets 2904C and 2904D, namely, places 220C and 220D, respectively.
Once each clone has reached its respective destination, property "tickets" of send frame 2902 (Figure 30) is a list of exactly one ticket. For example, WO 95/02219 ~ PCTIUS94/07397 agents 150A-1, 150A-2 and 150A-3 (Figure 29E) include send frames 2902, 2902-2 and 2902-3, respectively, which each include a single ticket, namely, respective tickets 2904D, 2904B and 2904C. Thus, each send frame contains the single ticket defining the trip taken by the corresponding agent clone. The single ticket is used as described above to form a ticket stub which is produced as a result by performance of operation "send" by each respective agent clone.
Thus, substantial savings are realized in computer storage space and data transportation time by deferring cloning of an agent performing operation "send" as long as two or more agent clones are taking trips which are coextensive in part. As described above in the context of operation "go", if an exception is thrown during performance of operation "go" by an agent, the agent can be placed in purgatory. Similarly, if an exception is thrown during performance of operation "send" by an agent clone, the agent clone can be placed in purgatory.
Additionally, if transportation of multiple clones of an agent which are represented by a single encoded agent, e.g. agent 150A-1 (Figure 29B), fails, each clone represented by the encoded agent is decoded, activated, and placed in purgatory as described above.
Interaction Between Accents: Operation "Meet"
Two agents, which occupy the same place, are capable of interacting with one another by one agent requesting that the other perform an operation, or set or query an attribute. Information is exchanged between the two agents through the arguments and result of the operation requested or through the attribute which is set or queried. For example, agent 150A and 150B occupy place 220B which is executing within engine 132B (Figure 15E).
Therefore, agent 150A and agent 150B are capable of interacting by (i) agent 150A requesting that agent 150B

WO 95/02219 PCT/US94/07397 __ :14367 ~
perform a feature or by (ii) agent 1508 requesting that agent 150A perform a feature.
Agent 150A is capable of requesting that agent 1508 perform a feature only if agent 150A contains a reference to agent 1508. Similarly, agent 1508 is capable of requesting that agent 150A perform a feature only if agent 1508 contains a reference to agent 150A. Agent 150A
obtains a reference to agent 1508 by requesting performance of operation "meet" by a meeting place of which agents 150A and 1508 are occupants. To this point, place 2208 has only been described as a place occupied by agents 150A and 1508. For purposes of the following discussion, place 2208 is a meeting place, i.e., a member of class "Meeting Place". Meeting places are discussed in more detail in Appendix A.
Figures 31A and 318 illustrate the interface of operation "meet". Frame 3100 is part of the execution state of agent 150A. Frame 3100 records the state of the performance of operation "meet" by meeting place 2208.
Meeting place 2208 is identified within the execution state of agent 150A as the responder of operation "meet".
Frame 3100 includes stack 3102, which is the current stack. Immediately prior to performance of operation "meet" (Figure 31A), stack 3102 contains, at its top, petition 3106. Petition 3106 is an argument consumed by performance of operation "meet". Petitions, i.e. members of class "Petition", are discussed more completely below and in Appendix A.
Logic flow diagram 3200 (Figure 32) shows the implementation of operation "meet" as performed by meeting place 220B. Engine 1328 (Figure 15E), in carrying out operation "meet" as performed by meeting place 2208, determines the petitioned agent by parsing petition 3106 (Figure 31A). Petition 3106 specifies a petitioned agent by name, by class or by both. A petitioned agent is an WO 95/02219 ~ PCT/US94/07397 - ~143~'~~
agent specified by a petition as the agent with which an agent requesting a meeting is configured to meet.
Engine 132B pops petition 3106 from the current stack in a pop petition step 3202 (Figure 32). Processing ' 5 transfers from pop petition step 3202 to a step 3204. In step 3204, engine 132B creates a new, empty list of telenames. The telenames of the list created in step 3204 are telenames of petitioned agents which have rejected a meeting with the requesting agent. The list is initially empty as initially no petitioned agent has rejected a meeting with the requesting agent.
Processing transfers from step 3204 to a find petitioned agent step 3206 in which the engine carrying out performance of operation "meet", e.g., engine 132B
(Figure 15E), finds a petitioned agent, i.e., an agent which satisfies petition 3106 (Figure 31A), whose telename is not an item of the list of telenames created in step 3204 (Figure 32). An agent satisfies petition 3106 (Figure 31A) as follows.
As described in greater detail in Appendix A, petition 3106 includes a property "agentName" and a property "agentClass" (neither shown). Both properties "agentName" and "agentClass" are optional and can therefore each be a nil. If property "agentName" of petition 3106 is a telename and property "agentClass" is nil, a petitioned agent is an agent whose telename is specified by the telename that is property "agentName" of petition 3106. If property "agentClass" is a citation and property "agentName" is nil, a petitioned agent is an agent which is a member of a class specified by the citation that is property "agentClass" of petition 3106.
If neither property is nil, a petitioned agent is an agent which satisfies both criteria. If both properties are nil, an exception of class "Meeting Invalid" is thrown, causing operation "meet" to fail.

~1~36'~ ~ -Processing transfers from find petitioned agent step 3206 to test step 3208 in which engine 132B determines whether a petitioned agent is found in find petition agent step 3206. If no petitioned agent is found in find petitioned agent step 3206, processing transfers from test step 3208 to a wait step 3222 which is described below.
Conversely, if a petitioned agent is found in find petitioned agent step 3206, processing transfers from test step 3208 to a meeting step 3210.
In meeting step 3210, the petitioned agent is requested to perform operation "meeting" by engine 132B
which is interpreting meeting place 220B. In performing operation "meeting", the petitioned agent agrees or refuses to participate in a meeting with agent 150A.
Operation "meeting" is discussed further below and in Appendix A.
After performance of operation "meeting" by the petitioned agent, processing transfers from meeting step 3210 to test step 3212. In test step 3212, engine 132B determines whether operation "meeting" completed successfully, indicating that the petitioned agent agrees to the meeting. If the petitioned agent refuses the meeting, i.e., processing transfers from test step 3212 to a first fully qualified test step 3234, which is described below. Conversely, if the petitioned agent agrees to the meeting, i.e., if operation "meeting" completed successfully, processing transfers from test step 3212 to a build contact step 3216.
In build contact step 3216, a contact 3108 (Figure 31B), whose subject is the petitioned agent, is created. Processing transfers from build contact step 3216 (Figure 32) to a push contact step 3218 in which contact 3108 (Figure 31B) is pushed on to stack 3102, thereby producing contact 3108 as a result. Processing transfers from push contact step 3218 to terminal - ~1~3~~2 step 3220 in which operation "meet" completes successfully.
However, as discussed above, if the petitioned agent throws an exception in performance of operation "meeting", thereby refusing to meet with agent 150A, processing transfers from test step 3212 to first fully qualified test step 3234. In first fully qualified test step 3234, engine 132B determines whether petition 3106 (Figure 31A) is fully qualified. Petition 3106 is fully qualified if the telename that is property "agentName" of petition 3106 is fully qualified. A telename includes both a property "authority" and a property "identity". Telenames are described more completely in Appendix A. Property "identity" of a telename is optional, i.e., can be a nil.
If property "identity" of a telename is nil, the telename is partially qualified and denotes all named objects of the authority that is property "authority" of the telename. On the other hand, if property "identity" of a telename is not nil, the telename is fully qualified and denotes exactly zero or one named object.
In first fully qualified test step 3234 (Figure 32), engine 132B determines whether petition 3106 (Figure 31A) is fully qualified. If petition 3106 (Figure 31A) is fully qualified, processing transfers from first fully qualified test step 3234 to terminal step 3236 in which ~n exception of class "Meeting Denied" is thrown, causing operation "meet" to fail. Thus, if petition 3106 (Figure 31A) is fully qualified and the one agent which satisfies petition 3106 rejects the meeting, operation "meet" fails.
On the other hand, if petition 3106 is not fully qualified, processing transfers from first fully qualified test step 3234 to an add agent to list step 3214.
In add agent to list step 3214, the petitioned agent found in find petitioned agent step 3206 is added to the list of telenames created in step 3204. Processing WO 95/02219 ~ i 4 3 6 '7 2 PCT/US94/07397 _.
transfers from add agent to list step 3214 to find petitioned agent step 3206.
As discussed above, if no petitioned agent is found in find petitioned agent step 3206, processing transfers from test step 3208 to wait step 3222. In wait step 3222, interpretation of agent 150A is suspended until petition 3106 (Figure 31A) expires or until an agent enters meeting place 220B, i.e., until meeting place 220B successfully performs operation "entering". As described more completely in Appendix A, petition 3106 includes a property "maximumWait" which defines a maximum amount of time that can elapse from the start of performance of operation "meet" before operation "meet" must conclude, either successfully or otherwise. Petition 3106 expires when the amount of time specified in property "maximumWait" has passed since the start of performance of operation "wait".
When petition 3106 expires or when an agent enters meeting place 220B, interpretation of agent 150A resumes and processing transfers from wait step 3222 (Figure 32) to a timeout test step 3224. In timeout test step 3224, engine 132B (Figure 15E) determines whether resumption of interpretation of agent 150A (Figure 31A) results from expiration of petition 3106. If interpretation of agent 150A is resumed as a result of the expiration of petition 3106 (Figure 31A), processing transfers from timeout test step 3224 (Figure 32) to terminal step 3226 in which an exception of class "Petition Expired" is thrown, causing operation "meet" to fail.
On the other hand, if interpretation of agent 150A
(Figure 31A) is resumed as a result of an agent entering place 220B, processing transfers from timeout step 3224 to test step 3228. In test step 3228, engine 132B (Figure 15E) determines whether the entering agent (not shown) is petitioned, i.e., satisfies petition 3106 (Figure 31A), and is not on the list of telenames created in step 3204 -126=

(Figure 32). If the entering agent does not satisfy petition 3106 (Figure 31A) or if the telename of the entering agent is an item of the list of telenames created in step 3204 (Figure 32), processing transfers from test step 3228 to wait step 3222.
If, on the other hand, the entering agent satisfies petition 3106 (Figure 31A) and the telename of the entering agent is not an item of the list of telenames created in step 3204 (Figure 32), processing transfers from test step 3228 to a second meeting step 3230. In second meeting step 3230, the entering agent is directed to perform operation "meeting". Processing transfers from second meeting step 3230 to test step 3232 in which engine 132B (Figure 15E) determines whether performance of operation "meeting" by the entering agent succeeded, the entering agent thereby accepting a meeting with agent 150A, or failed, the entering agent thereby rejecting a meeting with agent 150A.
If performance of operation "meeting" by the entering agent succeeds, processing transfers from test step 3232 to build contact step 3216. As described above,~in build contact step 3216 and the steps that follow, a contact to the petitioned, i.e., entering, agent is built, the contact is pushed on to the current stack, and operation "meet" completes successfully. If, on the other hand, performance of operation "meeting" by the entering agent throws an exception, processing transfers from test step 3232 to a second fully qualified test step 3238.
In second fully qualified test step 3238, engine 132B
(Figure 15E) determines whether petition 3106 (Figure 31A) is fully qualified. If petition 3106 is fully qualified, processing transfers from second fully qualified test step 3238 (Figure 32) to terminal step 3240 in which an exception of class "Meeting Denied" is thrown, causing operation "meet" as performed by meeting place 220B to fail.

~z1~3~'~ 2 Conversely, if petition 3106 (Figure 31A) is not fully qualified, processing transfers from second fully qualified test step 3238 to a second add agent to list step 3242. In second add agent to list step 3242, the telename of the petitioned, i.e., entering, agent is added to the list of telenames created in step 3204. Processing transfers from second add agent to list step 3242 to wait step 3222 which is described above.
Thus, a meeting is arranged between agent 150A and a petitioned agent (i) which is of the name and class specified by petition 3106, (ii) which occupies the responding meeting place, and (iii) which agrees to the meeting within the maximum time period specified in petition 3106. The last condition permits a first agent to request a meeting with a second agent which does not occupy the meeting place occupied by the first agent but which is expected to arrive at the meeting place within a certain amount of time.
Immediately following performance of operation "meet"
(Figure 31B), stack 3102 contains, at its top, contact 3108, as described above. Contact 3108 includes a property "subject" which is a reference to agent 150B
(Figure 15E). Property "subject" of contact 3108 (Figure 31B) is produced by querying attribute "subject" of contact 3108. Attribute "subject" of a contact is described in greater detail in Appendix A. Thus, agent 150A obtains a reference to agent 150B (Figure 15E) and is therefore capable of requesting that agent 150B
perform features.
As discussed below in the context of operation "meeting", agent 150B consumes, as an argument in performance of operation "meeting", a contact which identifies agent 150A as the agent requesting a meeting.
After performance of operation "meet" by meeting place 220B, the contact consumed by agent 150B includes a property "subject" which is a reference to agent 150A.

~I436'~~
Thus, agent 1508 obtains a reference to agent 150A and is capable of requesting that agent 150A perform features.
An agent, which is contacted, is a member of a class which inherits from mix-in class "Contacted". As described more fully in Appendix A below, mix-in class "Contacted" defines attribute "contacts" which provides access to property "contacts". Property "contacts" is a set of contacts. Agent 1508 (Figure 33) is a contacted agent and therefore includes property "contacts", which is set 3302. After performance of operation "meet" by meeting place 2208 as described above, contact 3404, which is the contact consumed in performance of operation "meeting" by agent 1508, is added to set 3302 by engine 1328.
Consent of the Petitioned Accent: Operation "MeetincL
As discussed above, a petitioned agent agrees or refuses to meet with an agent requesting a meeting by performance of operation "meeting" by the petitioned agent. Operation "meeting" is not defined or inherited by class "Agent". Instead, operation "meeting" is defined by mix-in class "Petitioned". Class "Agent" is not a subclass of mix-in class "Petitioned", but subclasses of class "Agent", which are subsequently defined by users of the present invention, can be subclasses of mix-in class "Petitioned". Thus, only agents which are members of such user-defined subclasses can perform operation "meeting"
and can therefore participate in meetings with other agents.
Figure 34A shows the execution state of agent 150A
immediately prior to performance of operation "meeting" by agent 1508. Frame 3400 is a part of the execution state of agent 150A and records the dynamic state of operation "meeting" as performed by agent 1508. Operation "meeting"
is requested by engine 1328; the performance of "meeting"
is recorded in the execution state of agent 150A. The ~~1436'~ 2 permit of either agent 150A or agent 1508, perferably agent 1508, may be debited by performance of operation "meeting". Agent 1508 (Figure 15E) is identified within engine 1328 as the responder of operation "meeting".
Stack 3402 is property "stack" of frame 3400 and is the current stack. Immediately prior to performance of operation "meeting", stack 3402 contains, from top to bottom, contact 3404 and petition 3406.
Contact 3404 identifies agent 150A as the agent requesting a meeting. Property "subjectName" (not shown) of contact 3404 is a telename which is equivalent to the telename of agent 150A, thereby identifying agent 150A as the requesting agent. Property "subjectClass" (not shown) of contact 3404 is a citation identifying the class of which agent 150A is an instance. Property "subject" (not shown) of contact 3404, which is normally a reference to agent 150A, is made nil by engine 132B. Thus, agent 1508, by consuming contact 3404 in performance of operation "meeting", has all the information necessary to properly identify agent 150A as the agent requesting the meeting, but has no reference to agent 150A and therefore has no way to interact with agent 150A. In this way, neither agent can interact with the other until both have agreed to the meeting.
Petition 3406 is a copy of petition 3106 (Figure 31A) supplied by agent 150A (Figure 15E) in requesting the meeting, as this argument is passed "byCopy". Passing an argument "byCopy" is discussed in greater detail below and in Appendix A.
Agent 1508 agrees to the meeting defined by petition 3406 (Figure 34A) with the agent identified by contact 3404 by performing operation "meeting" successfully. The meeting is refused by throwing an exception, thereby causing operation "meeting" to fail. The method for operation "meeting", as defined by mix-in class "Petitioned", never succeeds and always throws an ~I43~'~~
exception of class "Meeting Denied". However, subsequently defined subclasses of class "Agent", which inherit from mix-in class "Petitioned" can redefine the implementation of operation "meeting" so as to succeed ' 5 under certain circumstances.
Since agents are anticipated to serve a wide variety of needs and to perform a wide variety of services, the number of subclasses of class "Agent" and the variations of methods for operation "meeting" are quite large. For example, logic flow diagram 3500 (Figure 35) shows one such method for operation "meeting". In step 3502, contact 3404 (Figure 34A) and petition 3406 are popped from stack 3402. Processing transfers from step 3502 (Figure 35) to step 3504 in which attribute "subjectClass"
of contact 3404 (Figure 34A) is queried, thereby producing property "subjectClass" of contact 3404. Processing transfers from step 3504 (Figure 35) to a test step 3506, in which property "subjectClass" of contact 3404 (Figure 34A) is compared to the citation of a specific class. If property "subjectClass" of contact 3404 is equal to the citation, processing transfers from test step 3506 (Figure 35) to terminal step 3508. In terminal step 3508, operation "meeting" completes successfully and the meeting is thereby agreed to. Otherwise, if property "subjectClass" of contact 3404 (Figure 34) is not equal to the citation, processing transfers from test step 3506 (Figure 35) to terminal step 3510 in which an exception of class "Meeting Denied" is thrown causing operation "meeting" to fail, thereby rejecting a meeting with the agent identified by contact 3404 (Figure 34A).
As discussed above, contact 3404, which identifies agent 150A and is supplied to agent 150B (Figure 15E) as an argument in requesting performance of operation "meeting", contains no reference to agent 150A. Instead, property "subject" of contact 3404 (Figure 34A) is nil.
Once a meeting is successfully arranged by performance of ~1436'~2 operation "meet" by meeting place 220B (Figure 15E) as described above, engine 132B modifies property "subject"
of contact 3404 (Figure 34A) to be a reference to agent 150A.
Thus, the full generality and data processing capabilities of the computer instruction set described in Appendix A can be used to apply sophisticated logic, processing substantial amounts of data, to determine with which agents and under what circumstances specific agents are configured to agree to a meeting.
Figure 34B shows the state of frame 3400 immediately following performance of operation "meeting" by agent 150B
(Figure 15E). Stack 3402 (Figure 34B) is empty as operation "meeting" produces no result. The access of operation "meeting" is "system"; only an engine is capable of requesting operation "meeting". It should be noted that operation "meeting" defined by mix-in class "Petitioned" is distinct and separate from operation "meet" defined by class "Meeting Place" which is discussed above and in Appendix A.
Terminating Interaction Between Agents: Operation "Part"
As described above, two agents exchange references to each other and to objects owned by either agent during the course of a meeting between the two agents. Terminating interaction between the agents requires ensuring that neither agent contains a reference to the other agent or references to objects owned by the other agent. Either agent participating in a meeting between two agents occupying a meeting place can terminate the meeting by requesting performance of operation "part" by the meeting place. Agents 150A and 150B which occupy meeting place 220B and which are interpreted by engine 132B (Figure 15E), are shown in Figure 36. Agent 150A has obtained reference 150A-R1 to agent 150B, and agent 150B has obtained reference 150B-R1 to agent 150A, by the _ WO 95/02219 PCT/US94107397 ~1~3~72 successful performance of operation "meet" by meeting place 2208. Object 3602 is owned by agent 150A.
Similarly, object 3604 is owned by agent 1508. Figure 36 shows the state of agents 150A and 1508 after agents 150A
S and 1508 have interacted and exchanged references to objects 3602 and 3604. By giving to agent 1508 a reference to object 3602, agent 150A grants agent 1508 access to object 3602. Such access by references I50B-R1 and 1508-R2 enables agent 1508 to issue instructions to engine 1328 which cause agent 150A and object 3602, respectively, to take action in accordance with the instructions issued. Similarly, agent 1508 gives to agent 150A reference 150A-R2 to object 3604, thereby granting to agent 150A similar access to object 3604.
Agent 150A contains references 150A-R1, 150A-R2 and 150A-R3 to agent 1508, object 3604, and object 3602, respectively. References 150A-R1, 150A-R2 and 150A-R3 are (i) contained either within a stack or within a list of variables (neither shown) of a frame (also not shown) which is part of the execution state of agent 150A or (ii) stored as a property, or as a component of a property, of agent 150A. The execution state of an agent is discussed in detail below. Agent 1508 similarly contains references 1508-R1, 1508-R2 and 1508-R3 to agent 150A, object 3602, and object 3604, respectively. By giving to agent 150A reference 150A-R2 to object 3604, agent 1508 allows agent 150A (i) to request that object 3604 perform an operation or (ii) to copy a portion or all of object 3604.
Either agent 150A or agent 1508 is capable of terminating the interaction between agents 150A and 1508 by requesting that meeting place 2208 perform operation "part". For illustration purposes, agent 150A is the agent which requests that meeting place 2208 perform operation "part". Figures 37A and 378 illustrate the state of agent 150A immediately prior to and following, ~~1~3~~~ ~
respectively, performance of operation "part" by meeting place 2208. Frame 3702 is a part of the execution state of agent 150A and records the state of operation "part" as performed by meeting place 220B. Meeting place 2208 is property "responder" of frame 3702 and is therefore the responder of operation "part". Stack 3704 is property "stack" of frame 3702, which is also part of the execution state of agent 150A, and is the current stack.
At the top of stack 3704 is contact 3706.
Contact 3706 identifies agent 1508 as the agent from which agent 150A is to part. Engine 1328 (Figure 15E), in carrying out performance of operation "part" on behalf of meeting place 2208, directs agent 1508 to perform operation "parting". Agent 1508 consumes as the sole argument a contact identifying agent 150A as the agent with whom agent 1508 is parting. Property "subject" of the contact, which is usually a reference to agent 150A is voided by engine 1328 prior to requesting operation "parting" so that agent 1508 is not left with a reference to agent 150A after the meeting is terminated.
In performing operation "parting", agent 1508 produces no result. The dynamic state of performance of operation "parting" is not part of the execution state of either agent 1508 (Figure 15E) or agent 150A, but is instead part of the execution state of an engine process.
Any exception thrown by performance of operation "parting"
is not experienced by, and therefore has no effect upon, either agent 1508 or agent 150A. In performing operation "parting", agent 1508 is notified that agent 150A has terminated the meeting between agents 150A and 1508. If agent 1508 is a contacted agent, the item of property "contacts" of agent 150A that is the contact referencing agent 150A is removed from property "contacts" of agent 150B. The access of operation "parting" is "system"; therefore, only engine 132B can request that agent 1508 perform operation "parting".

z1~3~~2 . ..
Figure 378 illustrates the state of agent 150A
immediately following performance of operation "part" by meeting place 2208. Stack 3704 is empty as operation "part" produces no result.
Terminating a meeting between agents 150A and 150B
voids all references in agent 150A to agent 1508 and all objects owned by agent 1508. For example, in Figure 38, reference 150A-R1 that previously identified agent-1508 and reference 150A-R2 that previously identified object 3604 are voided. Similarly, reference 1508-R1 that previously referenced agent 150A and reference 1508-R2 that previously referenced object 3602 are voided within agent 1508 as well. Thus, agents 150A and 1508 are no longer capable of exchanging information. If agents 150A
and 1508 are contacted agents, a contact identifying agent 150A is removed from property "contacts" of agent 1508, and a contact identifying agent 1508 is removed from property "contacts" of agent 150A.
Applicabilitv of the Present Invention to Larae, Wide-Area Networks Thus far, a communications system is disclosed in which agent 150A (Figure 15A) transports itself, or a clone of itself, to meeting place 2208, which is occupied by agent 1508, by performance of operation "go" or operation "send", respectively. Agent 150A then requests performance of operation "meet" by meeting place 2208 to arrange a meeting between agent 150A and agent 1508.
During the meeting, agents 150A and 1508 are capable of exchanging information as described in greater detail below.
While the foregoing discussion pertains to a very simple use of the present invention, much more general uses can be made of the present invention. Through application of the set of computer instructions described herein and in Appendix A, an agent is capable of applying sophisticated and complex logic to determine to what places to transport itself and with which other agents and places to meet and exchange information.
Additionally, the present invention is not limited to networks of two or three computer systems, as described herein, and as previously stated, the present invention is not limited to homogenous networks. The set of computer instructions described herein and in Appendix A can be used to create and position places throughout a large area network such that agents can be created which travel throughout the large area network. As the disclosed instruction set can be implemented in heterogenous networks, the wide area network can include a wide variety of computer systems including large, mainframe computers;
local area networks; and small personal computers through which an agent in the wide area network can travel.
Furthermore, while a meeting is defined as an interaction between two agents, an agent is capable of participating in any number of meetings simultaneously and is therefore capable of interacting with a multitude of agents simultaneously, as long as those agents occupy the same meeting place occupied by the first-mentioned agent.
Interaction between Accents: Introduction to the Execution Model During a meeting between agent 150A and agent 150B, agent 150A (Figure 15E) interacts with agent 150B by issuing an instruction directing agent 150B to perform a feature. The following describes the mechanism by which performance of a feature is requested; this mechanism is discussed in greater detail below and in Appendix A.
An execution state is associated with each process.
As discussed above in the Glossary of Terms, each method has a dynamic state during performance of the method. The execution state of a process includes one or more frames, each frame recording the dynamic state of a method which ~1436'~2 is part of the execution state of the process. Within each frame is a stack on to which arguments are pushed . prior to requesting operations, and from which results are popped after performance of operations. The stack of the current frame is the current stack. The current frame is the frame which contains the computer instruction whose execution is currently being carried out by an engine.
Prior to requesting an operation, the requester pushes on to the current stack references to zero or more objects which are thereby supplied as arguments for the operation. The requester then pushes on to the current stack a reference to the responder. The responder is the object which is directed by the requester to perform the requested operation. Finally, the requestor requests the execution of an identifier which identifies the requested operation.
In carrying out performance of the requested operation, an engine pops an object from the top of the current stack, and the object is made the responder of the operation. A method implementing the operation is found within the classes of which the responder is a member and performed. In one embodiment, performance of the method (i) forms a new stack, (ii) pops the arguments of the operation from the current stack and pushes them on to the new stack, (iii) makes the new stack current, (iv) pops arguments from the new current stack, (v) pushes a result, if one is produced by the operation, on to the new current stack, and (vi) pops the result, if any, from the new current stack and pushes the result, if any, on to the previously current stack. The execution of operations, and features in general, is discussed in greater detail below and in Appendix A.
The following example illustrates the mechanism by which operations are requested. As discussed above, during the course of a meeting, agent 150A interacts with agent 150B by directing agent 150B to perform an operation. Agent 150A grants agent 1508 access to objects contained within agent 150A by supplying those objects as arguments to the operation requested. Agent 1508 grants agent 150A access to an object contained within agent 1508 by producing that object as a result of the requested operation.
Figures 39A-39F illustrate an example of agent 150A
(Figure 15E, not shown in Figures 39A-39F) interacting with and passing information to agent 1508.
Figure 39A shows empty stack 3902. Stack 3902 is the current stack of the execution state of agent 150A (not shown). While Figure 39A shows stack 3902 as being empty, such need not be the case. However, no objects on stack 3902 in Figure 39A are affected by this example.
String 3904 whose text is "this is a message" is pushed on to stack 3902 (Figure 398). A reference to agent 1508 is pushed on to stack 3902 (Figure 39C).
Agent 150A next executes an identifier whose text is "store" and which references an operation "store". In execution of the identifier, an object, i.e. agent 1508, is popped from stack 3902 and the object is directed to perform operation "store" (Figure 39D). In this example, operation "store" is defined for a class of which agent 1508 is a member and whose method is represented by logic flow diagram 4000 (Figure 40).
In performing operation "store", agent 1508 pops string 3904 from stack 3902 in step 4002 (Figure 40).
Although agent 1508 is described as performing operation "store", the frame (not shown) which records the dynamic state of agent 150B's performance of operation "store" is part of the execution state of agent 150A (not shown).
The charges allowance, i.e., property "charges", of the permit of agent 150A is therefore debited for processing resulting from performance of operation "store" by agent 1508. Agent 1508 is the object performing operation "store" since (i) the execution state of agent 150A (not ...
shown) identifies agent 150B as the responder and since (ii) the method of operation "store" is provided by a class of which agent 150B is a member and (iii) the internal state of agent 150B provides a context within which operation "store" is performed.
Processing transfers from step 4002 to step 4004 in which string 3904-C, which is a copy of string 3904, is created, resulting in the state shown in Figure 39E.
Strings 3904 and 3904-C are contained within frame 3906 which is part of the execution state of agent 150A and which records the dynamic state of operation "store" as performed by agent 150B. Processing transfers from step 4004 to step 4006 (Figure 40) where agent 150B stores string 3904-C. Processing transfers from step 4006 to step 4008 in which agent 150B discards string 3904, which was popped from stack 3902. Processing transfers from step 4008 to terminal step 4010 in which operation "store"
completes successfully. Step 4008 results in the state represented by Figure 39F. Thus, agent 150A has successfully transferred information to agent 150B.
In this manner, agent 150A interacts with agent 150B
by agent 150A requesting agent 150B to perform an operation, and agent 150A conveys to agent 150B objects as arguments to or results from operations requested or performed, respectively. While agent 150B is described as performing an operation at the request of agent 150A, it is reiterated here that performance of a feature by an object is, in actuality, performance of the feature by an engine in the context of that object defined by that object's internal state and class. The behavior of a particular feature can vary depending on the internal state of the object performing the feature, where the internal state of the object is defined in part by the object's properties.
Agent 150B can similarly transfer an object to agent 150A by pushing the object on to stack 3902, e.g., after step 4008 and before terminal step 4010. The object is popped from stack 4010 as a result of operation "store"
by agent 150A. Furthermore, since agent 150B contains a reference to agent 150A, as discussed above with respect to operation "meet", agent 150B is similarly capable of directing agent 150A to perform an operation and of supplying to agent 150A objects as arguments to the operation.
Portability of the Computer Instruction Set of the Present Invention A number of aspects of the present invention permit processes a great deal of mobility and versatility.
First, the computer instruction set of the present invention is homogeneously implemented in a homogeneous or heterogenous network. Second, classes are objects of the computer instruction set so that classes can travel throughout the network with mobile processes. And third, the computer instruction set is interpreted.
Homocxeneity of the Computer Instruction Set As discussed above, an agent, which includes data and computer instructions, is capable of traveling from one computer system to another and of executing on whichever computer system the agent may be found. Furthermore, the network throughout which agents are capable of traveling can be either homogeneous or heterogeneous. In view of the mobility of agents, it is not always possible to determine exactly which computer system of the network executes each of the computer instructions forming the procedural portion of an agent. It is therefore important that each computer system of the network, throughout which agents can travel, supports the same computer instruction set.

~14~672 Classes Represented as Obiects in the Disclosed Computer Instruction Set An agent executing within a first computer system is capable of creating new classes of objects which did not previously exist. The agent can then travel to an engine which is executing within a second computer system and interpretation of the agent, involving processing of members of the newly created class, continues within the second computer system. In such a case, the engine executing within the second computer system has never before encountered the newly created class or any members of that class. Therefore, classes are made objects of the disclosed instruction set so that classes are capable of traveling with agents from one computer system to another.
Class Structure Several objects combine to define and represent a class: a class object, an identifier, and a class definition. Class objects, identifiers and class definitions are members of the respective classes "Class", "Identifier" and "Class Definition". Class objects, i.e., members of class "Class", are objects which represent a set of objects whose features have the same interface and implementation. The set of objects are the "instances" of the class. Class "Class" and class objects are discussed in greater detail below and in Appendix A.
Identifiers, i.e., members of class "Identifier", are used within an interface object or an implementation object to reference a class. It should be noted that identifiers reference other types of objects as well.
Interface objects and implementation objects are discussed below in greater detail. A first identifier referencing a first class within a first interface object or implementation object is distinct from identifiers referencing all other classes within the first interface object or implementation object. However, a second interface object or implementation object can use an identifier equivalent to the first identifier to reference a second class.
An interface object and an implementation object together define a number of features, i.e., operations and attributes. The distinction between an attribute, in which case an object is directed to provide or alter a portion of the object's internal state, and an operation, in which case an object is directed to perform a number of computer instructions, serves primarily to aid in conceptualization of the present invention. Conceptually, an attribute is a special case of an operation. For example, querying an attribute of an object directs the object to perform computer instructions which cause the object to provide information regarding the internal state of the object.
It is reiterated here that performance of a feature by an object is, in actuality, performance of the feature by an engine in the context of that object defined by that object's internal state and class. The behavior of a particular feature can vary depending on the internal state of the object performing the feature, where the internal state of the object is defined in part by the object's properties.
A single identifier can simultaneously identify two separate features, each feature defined by a different class. For example, operation "add" is defined by class "Number" differently than by class "Dictionary". In selecting an interface and implementation, the engine carrying out performance of a feature selects the feature definition and associated method object appropriate for the class of object performing the feature. In other words, the feature definition is retrieved from the classes of which the responder is a member. A concrete class defining a feature defines both the feature's interface and the feature's implementation. An abstract __ ~1~~6'~2 class which defines a feature defines the feature's interface and can define the feature's implementation.
The "interface" of a feature defines the arguments consumed in performance of the feature as well as the result produced, if any. Additionally, the exceptions which can be produced by a feature are defined by the feature's interface. The interfaces of the features defined by a class are collectively represented and defined by an interface object, i.e., a member of class "Interface", which is property "interface" of a class definition object. Class "Interface" is discussed below in more detail.
A feature's implementation is defined by the feature's "method." A feature's method defines actions to be taken, i.e. computer instructions to be executed, in performance of the feature. A method object, a member of class "Method", is generally a sequence of instructions whose execution constitutes performance of the feature.
The implementations of the features defined or inherited by a class are collectively represented and defined by an implementation object, i.e., a member of class "Implementation", which is property "implementation" of a class definition object. Class "Implementation" is discussed below in more detail.
Feature definitions are inherited from a class's superclasses. In other words, if a feature is defined for a particular class, that feature is also defined for that class's subclasses. However, a subclass is capable of redefining the implementation of the feature, superseding the implementation inherited from the superclass. However if the feature is "sealed" by a superclass of the subclass for which the feature is defined, the subclass cannot redefine the implementation inherited from the superclass.
The features defined by a class are defined by the class's class definition, a member of class "Class Definition". The class itself is represented by a class ~143~~~
object, a member of class "Class". The class is referenced by an identifier, a member of class "Identifier".
A class is created by first creating a class definition, the structure of which is discussed in Appendix A. The class definition is directed to perform operation "makeClasses" to create an associated class object. The syntax and behavior of operation "makeClasses" is discussed in more detail in Appendix A.
The particular form and structure of the formed class object is not essential so long as (i) the information contained within the class definition object is preserved and (ii) citation 4108 (Figure 41A), interface object 4110, and implementation object 4112 can be derived from the class object. Condition (ii) is necessary for the encoding of a class object as described in Appendix B.
The engine carrying out performance of operation "makeClasses" on behalf of class definition 4100 (Figure 41A), in one embodiment, forms class object 4102 (Figure 41B) by replicating in class object 4102 the properties of class definition 4100. After performance of operation "makeClasses", class definition 4100 is discarded. The properties of class definition 4100 are discussed in greater detail below. As the properties of class definition 4100 are replicated in class object 4102, class object 4102 contains all of the information contained in class definition object 4100 and can derive the properties of class definition object 4100.
In another embodiment, processing efficiencies of compiled computer processes are realized by replicating information in citation 4108, interface object 4110 and implementation object 4112 to form citation structure 4108C (Figure 41C), interface structure 4110C and implementation structure 4112C, respectively, within class object 4104. Computer processes which are formed of computer instructions which are compiled, rather than WO 95/02219 PCTlUS94/07397 ~1~3672 interpreted, are generally more efficient, i.e., perform a given task within a fewer number of CPU clock cycles, than computer processes whose instructions are interpreted. As discussed above, computer instructions are initially in human-intelligible source code. Before a computer instruction can be executed by a computer, the instruction must be translated into a computer-intelligible object code or machine code. As interpreted computer processes require such translation for each computer instruction executed, and as compiled computer processes do not require such translation during execution of a compiled computer process, compiled computer processes can be substantially more efficient than interpreted computer processes.
Citation structure 4108C, interface structure 4110C
and implementation structure 4112C are objects which are analogous to citation 4108, interface object 4110 and implementation object 4112 and which are formed of compiled computer instructions such the computer instructions which collectively form the C++ programming language. By replicating the structures of citation 4108, interface object 4110, and implementation object 4112 (Figure 41A) in class object 4104 (Figure 41C) as citation structure 4108C, interface structure 4110C, and implementation structure 4112C, respectively, information regarding the class represented by class object 4104 can be retrieved substantially more efficiently than by retrieving the information from citation 4108, interface object 4110 and implementation object 4112 of class object 4102 (Figure 41B). The information contained within citation structure 4108C (Figure 41C), interface structure 4110C, and implementation structure 4112C is substantially equivalent to the information contained within the corresponding properties of class definition 4100 (Figure 41A) .

~1~3~'~2 As the information contained within a class object is substantially equivalent to the information contained within the class definition from which the class object is derived, discussion of the structure of a class definition also describes the nature of the information contained within a class object.
Class Definition Objects The features defined by a class are defined in the class definition object which defines the class. The structure of class definition object 4100 is shown in Figure 41A. Class definition object 4100 includes properties "citation", "interface" and "implementation".
Citation 4108 is property "citation" of class definition object 4100 and identifies the class defined by class definition object 4100 and further identifies the authority, title and edition of the class. The structure of a citation is discussed in greater detail below and in Appendix A.
Interface object 4110 and implementation object 4112 are, respectively, properties "interface" and "implementation" of class definition object 4100.
Interface object 4110 defines the interfaces of the features defined by class definition object 4100, and implementation object 4112 defines the implementations of the features defined by class definition object 4100.
Interface ob ' a The structure of interface object 4110 is shown in Figure 42. The properties of interface object 4110 include properties "isAbstract", "classFeatures", "sealedClassFeatures", "instanceFeatures", "sealedInstanceFeatures", "vocabulary" and "superclasses".
Boolean 4208 is property "isAbstract" of interface object 4110. If boolean 4208 is "true", the class, whose WO 95/02219 ~ ~ ~ PCT/US94/07397 interface is interface object 4110, is abstract. The class is concrete otherwise.
Lexicon 4210 is property "classFeatures" of interface object 4110. Lexicon 4210 is a lexicon of the definitions of the "class features" of the class. The class features of a class are the features for which the class itself, i.e., the class represented by class definition object 4100 (Figure 41A), is the responder. The keys of lexicon 4210 (Figure 42), e.g., identifier 4210K, are identifiers which reference the class features. The values of lexicon 4210, e.g., feature definition 4210V, are feature definitions defining the class features.
Feature definitions are discussed below in more detail.
Set 4212 is property "sealedClassFeatures" of interface object 4110. Set 4212 is a set of identifiers, e.g., identifier 4212I, which reference the class features which are sealed. Sealed features are features that subclasses are prevented from adapting. If a feature is not sealed, a subclass of the class represented by class definition object 4100 (Figure 41A) can adapt the feature by redefining the feature's implementation.
Lexicon 4214 (Figure 42) is property "instanceFeatures" of interface object 4110. Lexicon 4214 is a lexicon of the definitions of the "instance features"
of the class. The instance features of a class are features for which members of the class are the responder.
The keys of lexicon 4214 (Figure 42) (no key or value of lexicon 4214 is shown) are identifiers referencing the instance features, and the values of lexicon 4214 are feature definitions defining the instance features.
Set 4216 is property "sealedInstanceFeatures" of interface object 4110. Set 4216 is a set of identifiers (not shown) which reference the instance features which are sealed.
Lexicon 4218 is property "vocabulary" of interface object 4110. Lexicon 4218 associates identifiers of user-defined classes which are referenced in interface object 4110 with citations of those classes. The keys of lexicon 4218 are identifiers, e.g., identifier 4218K, and the values of lexicon 4218 are corresponding citations, e.g., citation 4218V. Citations are discussed in greater detail below and in Appendix A. Property "vocabulary" of interface object 4110, i.e., lexicon 4218, is "the vocabulary" of interface object 4110 and of the class defined by class definition object 4100 (Figure 41A).
List 4220 (Figure 42) is property "superclasses" of interface object 4110. List 4220 is a list of identifiers, e.g., identifier 4220I, which reference classes. The classes referenced by identifiers of list 4220 are immediate interface superclasses of the class defined by class definition object 4100 (Figure 41A). All but the last item of list 4220 (Figure 42) must identify mix-in classes, and the last item of list 4220 must identify a flavor. The importance of the last item of list 4220 identifying a flavor is discussed below in more detail in conjunction with the selection of a method object in performing a feature and in conjunction with object initialization.
Feature Definitions A feature definition, i.e., a member of class "Feature", is an object which defines the interface of a particular feature. The properties of feature definition 4210V (Figure 43) include properties "exceptions" and "isPublic". Class "Feature" is abstract;
therefore, feature definition 4210V is a member, not an instance, of~class "Feature" and is either an operation definition or an attribute definition, both of which are described below and in Appendix A. Set 4302 is property "exceptions" of feature definition 4210V. Set 4302 is a set of identifiers, e.g., identifier 4302I, which each reference a class of which exceptions thrown by the WO 95/02219 ~ PCTIUS94/07397 feature defined are members. When the feature defined by feature definition 4210V throws an exception, the exception is verified to be a member of one of the classes referenced by an identifier of set 4302. If the exception is not a member of a class referenced by an identifier of set 4302, a member of class "Unexpected Exception" is thrown in place of the exception. Otherwise, the exception is thrown.
Boolean 4304 is property "isPublic" of feature definition 4210V. Boolean 4304 indicates whether the feature defined by feature definition 4210V is a public feature or a private feature, i.e. whether the access of the feature defined is "public" or "private". Each feature has an associated "access" which determines which objects and under what circumstances each feature can be properly requested. The access of each feature of this invention is disclosed in Appendices A and B.
If Boolean 4304 is "true", the feature is public. A
feature whose access is public can be requested by any object. If Boolean 4304 is "false", the feature is private. A feature whose access is private can only be requested by the responder, i.e., an object can only request the feature of itself. System features, i.e., features whose access is "system", are all predefined and therefore are not defined by feature definitions.
Predefined features are implemented within an engine and therefore are not represented by objects such as feature definitions.
Attribute Definitions A feature definition that defines an attribute is an attribute definition, i.e., a member of class "Attribute"
Therefore, class "Attribute" is a subclass of class "Feature". As an attribute definition is a feature definition, an attribute definition inherits the properties defined or inherited by class "Feature" as ~1~3~'~2 described above. In addition, attribute definition 4400 (Figure 44) includes properties "constraint" and "isSet".
Constraint 4402 is property "constraint" of attribute definition 4400. Constraint 4402 constrains objects representing the attribute defined by attribute definition 4400. As discussed below in greater detail, a constraint places restrictions on an object and can restrict (i) the class of which the object is a member, (ii) whether the object must be an instance of the class, (iii) whether the object can be a nil and (iv) how the object is passed between a feature's requester and responder.
Boolean 4404 is property "isSet" of attribute definition 4400. Boolean 4404 indicates whether the attribute defined by attribute definition 4400 can be set as well as queried. If boolean 4404 is "true", attribute definition 4400 defines an attribute which can be set.
Otherwise, the attribute defined by attribute definition 4400 can only be queried.
Operation Definitions As discussed above, a feature is either an attribute or an operation. Therefore, a second subclass of class "Feature" is class "Operation". An operation definition, i.e., a member of class "Operation", defines the interface of an operation.
Figure 45 shows the structure of operation definition 4500. As an operation definition is a feature definition, operation definition 4500 includes the properties defined or inherited by class "Feature" as described above. In addition, operation definition includes properties "arguments" and "result".
List 4502 is property "arguments" of operation definition 4500. Property "arguments" of an operation definition is optional, i.e., property "arguments" of an operation definition can be a nil rather than a list. If WO 95102219 ~ ~ ~ ~ PCT/US94/07397 property "arguments" is a nil, the operation defined by operation definition 4500 consumes a variable number of arguments. The performance of an operation whose arguments are variable in number is described in Appendix A. Conversely, if property "arguments" is list 4502, the number of arguments consumed by the operation defined by operation definition 4500 consumes a fixed number of arguments, the number being equal to the length of list 4502. If property "arguments" is an empty list, the operation consumes no arguments. The items of list 4502, e.g., constraint 4502I, are constraints which constrain the arguments of the operation defined by operation definition 4500. As discussed below with respect to Figure 46, a constraint constrains an object's class and passage.
Constraint 4504 is property "result" of operation definition 4500. Constraint 4504 constrains the result of the operation defined by operation definition 4500.
Property "result" of an operation definition is also optional. If property "result" is a nil rather than constraint 4504, then the operation defined by operation definition 4500 produces no result.
Constraints As discussed above, a constraint restricts the class and passage of an object. The structure of constraint 4402 is shown in Figure 46. The properties of constraint 4402 include properties "classId", "ofClass", "isInstance", "isOptional", and "passage".
Properties "ofClass" and "classId" are alternative ways for specifying the class of which the constrained object must be a member. Class object 4602 is property "ofClass" of constraint 4402. Class object 4602 represents the class of which an object upon which constraint 4402 is imposed must be a member. Property "ofClass" is optional and, therefore, can be a nil. If property "ofClass" is nil, the class of which the object must be a member is referenced by identifier 4610 which is property "classId" of constraint 4402.
Property "ofClass" is a class object and therefore can only be a class which exists at the time constraint 4402 is created. However, requiring that a property of a constraint is a class which exists prevents users of the present invention from defining multiple interdependent classes. For example, a constraint in a feature definition of a first class can refer to a second class and a second constraint in a second feature definition in the second class can refer to the first class. Providing property "classId" of constraint 4402 as an alternative to property "ofClass" enables users to define multiple interdependent classes simultaneously.
Boolean 4604 is property "isInstance" of constraint 4402. Boolean 4604 indicates whether an object upon which constraint 4402 is imposed must be an instance of the class represented by class object 4602. If boolean 4604 is "true", the constrained object must be an instance of the class. Otherwise, the constrained object's membership in the class is sufficient to satisfy constraint 4402.
Boolean 4606 is property "isOptional" of constraint 4402. Boolean 4606 indicates whether a nil object satisfies constraint 4402, properties "ofClass" and "classId" notwithstanding. If boolean 4606 is "true", a nil object satisfies constraint 4402. As illustrated in numerous examples herein, a nil object can be placed on a stack in lieu of an omitted optional argument or result of a feature.
Identifier 4610 is property "passage" of constraint 4402. Identifier 4610 indicates in what manner an object on which constraint 4402 is imposed is passed as a parameter, i.e., as an argument or result.
Identifier 4610 has four possible values: "byRef", WO 95/02219 ~ ~ ~ J ~ ~,~ ~ PCT/US94/07397 "byUnprotectedRef", "byProtectedRef" and "byCopy".
Setting identifier 4610 to a value other than one of these four throws an exception that is a member of class "Passage Invalid".
In passing a reference to an argument between the requestor of a feature and the responder of the feature, an engine takes a source reference to the argument from the requestor and provides a destination reference to the responder. In passing a reference to a result, the source reference is taken from the responder and a destination reference is provided to the requestor.
If the parameter is passed "byRef", the source reference and the destination reference are the same. If the parameter is passed "byUnprotectedRef", an exception of class "Reference Protected" is thrown if the source reference is a protected reference; otherwise, the destination reference is the source reference. If the parameter is passed "byProtectedRef", the destination reference is a protected reference to the object referenced by the source reference. If the parameter is passed "byCopy", the destination reference is an unprotected reference to a copy of the object referenced by the source reference. Passage of parameters is discussed further in Appendix A.
Thus, constraint 4402 (Figure 44) restricts the class and passage of an object.
Implementation Object As discussed above, implementation object 4112 (Figure 41A) defines the class implementation of the class defined by class definition object 4100. A class implementation defines the following: (i) zero or more properties of the members of the class defined; (ii) the respective implementations of the various class features and (iii) instance features defined by a class; (iv) methods for setting attributes defined for the class; (v) WO 95/02219 . PCT/US94I07397 ~~14~~7~
methods for converting objects from the class defined to a second class; (vi) methods for converting objects from a second class to the class defined; (vii) definitions of identifiers used to define classes; and (viii) the implementation superclasses of the class defined.
List 4702 (Figure 47) is property "properties" of implementation object 4112. List 4702 defines the properties that implementation object 4112 implements.
The items of lexicon 4702, e.g., identifier 4702I, are identifiers of the properties defined by implementation object 4702.
Lexicon 4704 (Figure 47) is property "classMethods"
of implementation object 4112. Lexicon 4704 provides methods for the class features of the class represented by class definition object 4100 (Figure 41A). The keys of lexicon 4704, e.g., identifier 4704K (Figure 47), are identifiers of the class features, and the values of lexicon 4704, e.g., method object 4704V, are method objects which define the implementation of the class features.
A method object, i.e., a member of class "Method", is an object which defines the method of a feature. The method of a feature is the particular series of steps taken by an engine, and therefore the computer system within which the engine is executing, in carrying out performance of the feature.
A method is defined in terms of the disclosed instruction set as the instruction set is amended by any user-defined classes or features. Method objects are discussed in greater detail below and in Appendix A.
Lexicon 4706 is property "instanceMethods" of implementation object 4112. Lexicon 4706 associates identifiers of instance features with corresponding method objects, which in turn define the implementation of the instance features. The structure of lexicon 4706 is similar to the structure of lexicon 4704.
Lexicon 4707 is property "setMethods" of implementation object 4112. Lexicon 4707 associates identifiers, which identify respective attributes, with method objects. Each of the method objects of lexicon 4707 implements the attribute of the corresponding identifier when executed in the presence of a "set" modifier. In other words, each method object of lexicon 4707 defines the implementation of the setting of a corresponding attribute.
The execution of an attribute in the presence of a "set"
modifier is discussed in Appendix A.
Lexicon 4708 is property "fromMethods" of implementation object 4112. Lexicon 4708 associates identifiers with method objects. Each identifier references a class and is associated with a method object which provides the implementation of the conversion from the referenced class to the class defined by class definition object 4100 (Figure 41A), i.e., the class whose implementation is implementation object 4112 (Figure 47).
Lexicon 4710 is property "toMethods" of implementation object 4112. Lexicon 4710 associates identifiers with method objects. Each identifier references a class and is associated with a method object which provides the implementation of the conversion to the referenced class from the class defined by class definition object 4100 (Figure 41A), i.e., the class whose implementation is implementation object 4112 (Figure 47).
The structure of lexicons 4707, 4708 and 4710 is similar to the structure of lexicon 4704.

Lexicon 4712 is property "vocabulary" of implementation object 4112. Lexicon 4712 defines the identifiers used by implementation object 4112 to identify classes. The keys of lexicon 4712, e.g., 155a identifier 4712K, are identifiers used within implementation object 4112 to identify specific user-defined classes. The values of lexicon 4712, e.g., citation 4712V, are corresponding citations which identify the class identified by the associated identifier.
Lexicon 4712 provides a translation between an identifier used to identify a user-defined class within implementation object 4112 and a citation used to identify the class throughout the network. Citations are discussed in greater detail below and in Appendix A.
List 4714 is property "superclasses" of implementation object 4112. The items of list 4714, e.g., identifier 4714I, are identifiers which identify the implementation superclasses of the class defined by class definition 4100 (Figure 41A). List 4714 (Figure 47), i.e.
property "superclasses" of implementation object 4112, can contain identifiers of classes which are not items of list 4220, i.e. property "superclasses" of interface object 4110. For example, a first class may benefit from the ability to use features of a second class in implementing features of the first class. However, in designing such a class, it is not always preferred that the features of the second class are themselves inherited by the first class. In such cases, the second class is made an implementation superclass but not an interface superclass of the first class.
Thus, implementation object 4112 defines, for the class defined by class definition object 4100 (Figure 41A), (a) properties of the members of the class defined, (b) the class and instance feature methods, (c) methods for setting attributes, (d) methods for conversion to and from the class defined, (e) identifiers used to identify classes, and (f) implementation superclasses.

WO 95/02219 ~ PCT/US94/07397 Method Ob'e As discussed above, a method object defines the computer instructions executed in performance of a feature or a conversion. Figure 48 shows the structure of method ' 5 object 4800. Properties of method object 4800 include properties "procedure" and "variables". Procedure 4802 is property "procedure" of method object 4800. Performance of procedure 4802 constitutes performance of the feature or a conversion implemented by method object 4800.
Procedures are discussed in more detail below.
List 4804 is property "variables" of method object 4800. List 4804 defines the variables that are used by procedure 4802. The items of list 4804, e.g., identifier 4804I, are identifiers which reference the variables defined for method object 4800. Associated objects which represent the values of the variables of a method are stored in a list as property "variables" of a user-defined frame during performance of the method.
Frames are discussed in greater detail below and in Appendix A.
Procedure The structure of procedure 4802 is shown in Figure 49.
Procedure 4802 includes list 4902 of objects 4902A-4902E.
List 4902 is not a member of class "List", but is instead an integral part of procedure 4802. In other words, list 4902 is not a property or item of procedure 4802 and no feature is provided which permits list 4902 to be treated as a member of class "List". In one embodiment of the present invention, class "Procedure" is an implementation subclass of class "List" but is not an interface subclass of class "List". Therefore, the features that are defined by class "List" are not inherited by class "Procedure" but the methods defined by ~1 g~36'~ 2 class "List" can be used by a procedure to create and access list 4902.
Objects 4902A-4902E are executed objects, i.e., members of mix-in class "Executed". Procedure 4802 is performed by executing objects 4902A-4902E sequentially, i.e., in the order of object 4902A, object 4902B, object 4902C, object 4902D and object 4902E. During execution of procedure 4802, a predefined frame recards and maintains the dynamic state of a performance of procedure 4802, including which of objects 4902A-4902E is executing. Predefined frames are discussed below and in detail in Appendix B.
Citation A citation is an object which identifies a series of objects each of which is backward- or forward-compatible with the others. A citation further identifies a particular object in the series and the process, or the authority of the process, that created the particular object. As discussed briefly above, classes are cited objects. Therefore, citations are used to identify series of classes which are backward- and forward-compatible with one another. Backward- and forward-compatibility are discussed in greater detail below and in Appendix A.
It should be noted that objects whose citations identify the objects as forward- or backward compatible with one another are not necessarily forward- or backward-compatible. For example, two class definition objects which are identified by their respective citations as compatible are merely intended to be compatible.
Citation 5000 (Figure 50) is a member of class "Citation". The properties of citation 5000 include properties "title", "majorEdition", "minorEdition" and "author". Property "title" of citation 5000 is identifier 5008. Identifier 5008 is therefore the title of citation 5000. The title of a citation references the ~1~3~'~2 series that the citation represents. The title of a citation is interpreted relative to the citation's authority. In other words, two citations of different authorities can use equivalent titles. The two citations are distinguished from one another by the different authorities. The authority of a citation is identified by property "author" which is discussed below.
Properties "majorEdition" and "minorEdition" a-re integers 5004 and 5006, respectively. Integers 5004 and 5006 are therefore the major and minor editions, respectively, of citation 5000. The relation between two objects in a series is determined by the relative major and minor editions of the objects. For example, objects 5102 and 5104 (Figure 51) are cited by citations 5106 and 5108, respectively. Integers 5110 and 5112 represent the major and minor editions, respectively, of citation 5106. Integers 5114 and 5116 represent the major and minor editions, respectively, of citation 5108.
If integer 5110, the major edition of citation 5106, is greater than integer 5114, the major edition of citation 5108, then object 5102 was created subsequent to the creation of object 5104 and is backward-compatible with object 5104. If object 5102 is backward-compatible with object 5104, object 5104 is forward-compatible with object 5102. Conversely, if integer 5110 is less than integer 5114, object 5104 was created subsequent to the creation of object 5102 and is backward-compatible with object 5102.
If integers 5110 and 5114 are equal, the relative values of integers 5112 ad 5116, respectively representing the minor editions of objects 5102 and 5104, determine the relation between objects 5102 and 5104. For example, if integer 5112 is greater than integer 5116, object 5102 was created subsequent to the creation of object 5104 and is backward-compatible with object 5104. Conversely, if integer 5112 is less than integer 5116, object 5104 was ~~~~~"~
created subsequent to the creation of object 5102 and is backward-compatible with object 5102.
Property "author" is telename 5002 (Figure 50).
Telename 5002 is therefore the author of citation 5000.
Telename 5002 is the name of the process creating the object identified by citation 5000. All objects of a single series are created by processes of a single authority.
While a more precise definition of "forward-" and "backward-compatibility" is given in Appendix A, the following example illustrates the meaning and utility of such concepts. As defined in Appendix A, a member of class "Integer", i.e., an integer, can perform a number of arithmetic operations, including "add", "subtract", "multiply" and "divide". Property "citation" of class "Integer" is a first citation identifying class "Integer".
In implementing the set of computer instructions described herein and in defining new classes and new features in using the set of computer instructions described herein, a plethora of features are designed in reliance of members of class "Integer" comporting with the description of class "Integer" as described in Appendix A.
Suppose in a subsequent version of the set of computer instructions described herein that a second class "Integer" is defined to provide an operationv"randomize"
to generate random numbers. The former class "Integer" is called the first class "Integer". Property "citation" of the second class "Integer" is a second citation. Suppose further that there are no other differences between the first class "Integer" and the second class "Integer" and that any feature requested of a member of the first class "Integer" is satisfied by a member of the second class "Integer". Thus, the features which are designed in reliance of integers comporting with the description of the first class "Integer", function properly given integers of the second class "Integer". The second class :143 ~'~~
"Integer" is therefore backward-compatible with the first class "Integer". Conversely, the first class "Integer" is forward-compatible with the second class "Integer".
Property "title" of both the first and the second citations is equal and references either class "Integer".
However, the major and minor editions of the first and second citations reflect the forward- and backward-compatibility relation between the first and second-classes "Integer". Permitting backward- and forward-compatible objects to co-exist within the network permits objects created by earlier versions of the set of computer instructions described herein to travel to and operate within engines implementing a hater version of the set of computer instructions described herein. It is clear from the foregoing that a class is both forward- and backward-compatible with itself. While the illustrative example above demonstrates forward- and backward-compatibility in the context of a predefined class, user-defined classes are typically changed more frequently than predefined classes. Therefore, providing for forward- and backward-compatibility among classes is particularly important in the context of user-defined classes.
Thus, classes are made part of the disclosed instruction set and are therefore portable and can travel with an agent from a first computer system to a second computer system.
Interpreted Instruction Set As discussed above, three aspects of the present invention make computer processes particularly mobile and general. The first is that the computer instruction set of the present invention is implemented homogeneously in a homogeneous or heterogeneous network. Secondly, classes are objects of the disclosed computer instruction set so that the instruction set is both extensible and mobile.
That is, classes, which are not defined on remote computer systems, are transported with mobile computer processes to such remote computer systems. The following discussion concerns the third aspect making the computer instruction set particularly general and mobile; the computer instructions of the disclosed computer instruction set are interpreted.
Most computer processes are first written in a human-readable source code which is then compiled into object code and linked into machine code, which is entirely unintelligible to human beings. The primary advantage of translating the source code into machine code is that the translation need occur only once as the machine code can be executed by a computer system's CPU any number of times without recompilation of the source code. However, two computer systems of a heterogeneous network can have incompatible CPUs which do not both execute identical machine code instructions. It is therefore preferred that agents, traveling from a first computer system to a second computer system, be represented in a standardized instruction set that is not specific to either the first or the second computer system.
Since the advantages of one-time compilation of source code are eliminated by the heterogeneity of the network, the disclosed computer instruction set is interpreted. The term "interpreted" is used herein as it is understood in the art; a computer instruction in a series of computer instructions is read, translated into machine code, and executed by an engine before the next computer instruction in the series is read.
Interpreting, rather than compiling, instructions of the disclosed instruction set provides greater generality.
A first agent can travel from a first computer system to a second computer system and meet with a second agent there which gives to the first agent a procedure. The first agent can then perform the procedure which the first agent was not originally designed to perform.

_ X143672 The following discusses the interpretation of the disclosed computer instructions. The discussion focuses on procedures and execution of items of procedures, the items being individual computer instructions.
Execution of a Procedure A procedure is performed by issuing an instruction requesting that the procedure perform operation "do".
Other operations which cause performance of a procedure are discussed below and in Appendix A. Performance of operation "do" as defined for class "Procedure" is shown by logic flow diagram 5200 (Figure 52A) and is represented by a predefined frame 5250 (Figure 52B). Predefined frame 5250 includes properties "position" and "procedure", which are, respectively, integer 5252 and procedure 4802.
Predefined frames are discussed in greater detail in Appendix B. Predefined frame 5250 records the dynamic state of operation "do" as performed by procedure 4802 (Figure 49).
In step 5202 (Figure 52A), integer 5252 is set to a value of one. Integer 5252 identifies the position, within procedure 4802, of the item whose execution was most recently begun. Processing transfers from step 5202 (Figure 52A) to end of procedure test step 5204 in which integer 5252 (Figure 52B) is compared to the length of procedure 4802. If integer 5252 is less than or equal to the length of procedure 4802, processing transfers from end of procedure test step 5204 (Figure 52A) to step 5208.
In step 5208, the item, whose position within procedure 4802 (Figure 52B) is integer 5252 is performed.
The performance of an executed object is discussed below in more detail and in Appendix A.
Once an item of procedure 4802 is performed, processing transfers from step 5208 (Figure 52A) to exception test step 5210. In exception test step 5210, the engine interpreting procedure 4802 (Figure 52B) WO 95/02219 PCT/US94/07397 _ determines whether performance of the item succeeded or threw an exception. If an exception is thrown, processing transfers from exception test step 5210 (Figure 52A) to terminal step 5212 in which operation "do" fails and the performance of procedure 4802 (Figure 52B) completes.
If no exception is thrown by performance of the item at the position within procedure 4802 indicated by integer 5252 in step 5208 (Figure 52A), processing transfers from exception test step 5210 to step 5214 in which integer 5252 (Figure 52B) is incremented. Processing transfers from step 5214 (Figure 52A) to end of procedure test step 5204 in which property "position" of predefined frame 5250 (Figure 52B), i.e., integer 5252, is compared to the length of responding procedure 4802 as described above.
If integer 5252 (Figure 52B) is greater than the length of procedure 4802, processing transfers from end of procedure test step 5204 (Figure 52A) to terminal step 5216 in which operation "do" completes successfully.
Thus, performance of operation "do" by procedure 4802 (Figures 49 and 52B) results in the execution of each of the items of procedure 4802 in the order described above.
Execution Model As discussed above; the primary function of an engine, e.g., engine 132B (Figure 15A), is the performance of procedures, i.e., the execution of the items of procedures. The execution model of the disclosed embodiment of the present invention is the process by which items of procedures are executed. Discussion of the execution model of the disclosed embodiment is divided into six sections. First, the structure and organization of frames is discussed. Frames are objects which record the dynamic state of a method implementing a feature or a conversion. Second, the execution of identifiers is discussed. The execution of an identifier which _ ~ PCT/US94/07397 references a feature causes the feature referenced to be performed. Third, the process by which a method object implementing a feature or a conversion is selected from the class hierarchy is discussed. The structure of class objects, from which a method object is retrieved, is described above. Fourth, the escalation of a feature is discussed. Fifth, the execution of modifiers is discussed. And sixth, the execution of selectors is discussed.
Frames A frame is an object which records the dynamic state of a method implementing a feature or a conversion during the feature's or conversion's execution. The execution state of a process, i.e., an agent or a place, is recorded in one or more frames. For example, the execution state of process 5300 (Figure 53) is shown. When a process is created, the p ~cess is asked to perform operation "live".
Created within process 5300 is a property "frames" which is stack 5302 which in turn represents the execution thread of process 5300. Stack 5302 contains the frames which collectively form the execution state of process 5300. When process 5300 is first asked to perform operation "live", the sole item on stack 5302 is user-defined frame 5304 which records the dynamic state of operation "live".
Since predefined classes "Process", "Agent" and "Place" are abstract, processes of the disclosed embodiment of the present invention are instances of user defined classes. Predefined abstract classes "Process", "Agent" and "Place" do not define an implementation for operation "live". Therefore, frame 5304 recording the dynamic state of the method of operation "live" is necessarily a user-defined frame. A user-defined frame is a frame which records the dynamic state of an operation with a user-defined implementation during execution of the WO95/02219~~~~~Z

operation. The structure of user-defined frame 5304 is shown in Figure 54.
The properties of user-defined frame 5304 (Figure 54) include properties "class", "responder", "procedure"
"position", "stack" and "variables". Class object 5402 is property "class" of user-defined frame 5304, which is a property inherited from superclass "Object". Class object 5402 represents the class of which user-defined frame 5304 is an instance and therefore represents class "User-defined Frame" or a subclass thereof.
Object 5404 is property "responder" of user-defined frame 5304. Object 5404 is the responder of the feature or conversion whose dynamic state user-defined frame 5304 records. In the context of Figure 53, object 5404 is process 5300 as process 5300 is the responder of operation "live".
Procedure 5406 (Figure 54) is property "procedure" of user-defined frame 5304, which is a property inherited from superclass "Procedure Frame". Procedure 5406 is property "procedure" of the method object whose dynamic state user-defined frame 5304 records. For example, if user-defined frame 5304 records the dynamic state of method object 4800 (Figure 48), procedures 4802 and 5406 are the same procedure.
Integer 5408 is property "position" of user-defined frame 5304 inherited from superclass "Procedure Frame".
Integer 5408 identifies the position within procedure 5406 of the item of procedure 5406 which is currently being executed. For example, if the second item of procedure 5406 is the item of procedure 5406 whose execution has begun and has not yet finished, the value of integer 5408 is two.
User-defined frame 5304 includes two additional properties, namely, properties "stack" and "variables".
Stack 5410 is property "stack" of user-defined frame 5304 and is the stack of user-defined frame 5304. At the beginning of execution of procedure 5406, stack 5410 contains the arguments consumed by performance of the feature or conversion whose dynamic state is recorded by user-defined frame 5304. After execution of procedure 5406, stack 5410 contains the result, if any, produced by performance of the feature or conversion.
List 5412 is property "variables" of user-defined frame 5304. The items of list 5412, e.g. object 5412I, are the variables of the frame. Variables record the dynamic state of user-defined frame 5304 as procedure 5406 is performed. Variables which are items of list 5412 (Figure 54) are referenced by like-positioned items of property "variables" of the method object whose dynamic state is recorded by user-defined frame 5304. For example, if user-defined frame 5304 records the dynamic state of method object 4800 (Figure 48), identifier 4804I
which is at position one within list 4804 references object 5412I (Figure 54) which is at position one within list 5412.
Thus, user-defined frame 5304 records the dynamic state of a user-defined feature or conversion by recording (i) the responder of the feature or conversion, (ii) the procedure implementing the feature or conversion, (iii) the position within the procedure of the currently executing instruction, (iv) the stack containing the arguments or result of the feature or conversion and (v) the variables which record the dynamic state of user-defined frame 5304.
Execution of Executed Obiects As discussed above, the performance of a procedure is the sequential execution of the items of the procedure.
In general, the execution of an executed object, i.e., an object which inherits from mix-in class "Executed", results in a reference to the object being pushed on to the stack, e.g., stack 5410, of the frame representing the ~14~~'~
execution state of the procedure. However, the execution of identifiers, modifiers and selectors are exceptions to this general rule.
Identifiers Unless executed in the presence of a modifier indicating otherwise, an executed identifier is presumed to reference a feature and therefore invokes an operation or queries an attribute. Execution of an identifier within a first procedure causes the feature referenced by the identifier to be performed, thereby causing the performance of a second procedure. The second procedure is the procedure of the method object which implements the operation or attribute referenced by the identifier. The execution of an identifier which references a feature is shown by logic flow diagram 5500 (Figure 55).
Logic flow diagram 5500 is discussed in the context of Figures 56A-56C. Stack 5302 (Figure 56A) is the execution thread, i.e., property "frames", of process 5300. User-defined frame 5304 is the current frame. The current frame is the frame containing the procedure, i.e., procedure 5406, whose items are currently being executed by an engine. In other words, the next executed object, which an engine executes in the interpretation of process 5300, is retrieved from the procedure of user-defined frame 5304.
Procedure 5406 is property "procedure" of user-defined frame 5304. Integer 5408 is property "position"
of user-defined frame 5304 and indicates the position within procedure 5406 of the item whose execution was most recently begun by the engine processing process 5300.
The execution of an identifier occurs when the item of procedure 5406 at the position indicated by integer 5408 is an identifier. In this example, the identifier references an operation. The operation referenced by the identifier is sometimes called "the WO 95/02219 ~ PCT/L1S94/07397 subject operation" in the context of Figures 56A-56C and 55. The execution of the identifier is shown by logic flow diagram 5500 (Figure 55).
The requester's stack, i.e. stack 5410, is produced by querying attribute "stack" of user-defined frame 5304 in step 5502. Since performance of procedure 5405 (Figure 56A) of user-defined frame 5304 is requesting execution of the identifier at the position within procedure 54Q6 indicated by integer 5408, property "responder" of user-defined frame 5304, which is the current frame, is the requester of the operation referenced by the identifier.
In the context of Figure 56A, the responder is process 5300. Stack 5410 is therefore the "requester's stack".
Processing transfers from step 5502 (Figure 55) to a stack empty test step 5504. At the time the identifier is executed, the object at the top of stack 5410 is the responder of the subject operation. In stack empty test step 5504, the engine interpreting process 5300 determines whether stack 5410 is empty. If stack 5410 is determined to be empty, processing transfers from stack empty test step 5504 to terminal step 5506 in which an exception of the class "Responder Missing",is thrown and the subject operation fails. Otherwise, if stack 5410 is not empty, processing transfers from stack empty test to step 5508 in which the object (not shown) at the top of stack 5410 is popped from stack 5410. The object popped from stack 5410 is the intended responder of the subject operation of the executing identifier and is sometimes referred to as "the responding object" in the context of Figures 56A-56C and 55.
It should be noted that the subject operation is user-defined and therefore represented by a user-defined frame. Features which are predefined are represented directly in an engine. In other words, the instructions which form a predefined operation are included in the instructions which form the computer process that is an ~143~'~2 engine. If the executed identifier references a predefined operation, the instructions which define the operation within the engine are located and performed.
Processing transfers from 5508 to step 5510. In step 5510, the operation definition defining the subject operation is retrieved. The selection of the operation definition and the associated method object is discussed below in detail. Processing transfers from step 55-10 to access test step 5512. In access test step 5512, the access of the subject operation is verified. Property "isPublic", e.g., Boolean 4404 (Figure 44), is retrieved from the operation definition. If the access of the subject operation is not "public", i.e., if property "isPublic" is "false", and the responding object is not property "responder" (not shown) of user-defined frame 5304, processing transfers from access test step 5512 to terminal step 5514.
In terminal step 5514, an exception of class "Feature Unavailable" is thrown and the subject operation fails.
Otherwise, if property "isPublic" is "true" or the responding object is property "responder" of user-defined frame 5304, the access of the subject operation is acceptable and processing transfers from access test step 5512 to number of arguments test step 5516.
In number of arguments test step 5516, property "arguments" of the operation definition is produced by querying attribute "arguments" of the operation definition. The number of items in the list of property "arguments" is compared to the number of objects on stack 5410. If property "arguments" defines a number of arguments that is greater than the number of objects present on stack 5410, processing transfers from number of arguments test step 5516 to terminal step~5518. In terminal step 5518, an exception of class "Argument Missing" is thrown, causing the subject operation to fail.

_ WO 95102219 PCT/US94/07397 ~I43~'~2 It should be noted that any exception thrown by execution of the identifier which references the subject operation, e.g., in either terminal step 5506 or terminal step 5514, causes the subject operation to fail, throwing the same exception.
If stack 5410 contains at least as many objects as there are arguments as indicated by property "arguments"
of the operation definition retrieved in step 5510, processing transfers from number of arguments test step 5516 to classes of arguments test step 5520. In classes of arguments test step 5520, the engine interpreting process 5300 determines whether each object on stack 5410 in the position of an argument satisfies the corresponding constraint within the list that is property "arguments" of the operation definition retrieved in step 5510. For example, constraint 4502I of list 4502 (Figure 45), which is property "arguments" of operation definition 4500, corresponds to the object at the top of stack 5410. The structure of a constraint is discussed in greater detail above and in Appendix A.
As discussed above in greater detail, a constraint restricts an object by specifying (i) a class of which the object must be a member, (ii) whether the object must also be an instance of the class specified, and (iii) whether a nil satisfies the constraint irrespective of the class specified. Each object on stack 5410 which corresponds to a constraint is examined to determine whether the object satisfies the constraint in classes of arguments test step 5520. If any of the objects on stack 5410 does not satisfy the corresponding constraint of property "arguments" of the operation definition, processing transfers from classes of arguments test step 5520 to terminal step 5522 in which an exception of class "Argument Invalid" is thrown.
Conversely, if each of the objects on stack 5410 satisfies the corresponding constraint, processing ~~1436"~~
transfers from class of arguments test step 5520 to step 5524. In step 5524, the method object defining the implementation of the subject operation in the context of the responding object is selected. The method object is selected from the classes of which the responding object is a member as discussed in greater detail below.
Processing transfers from step 5524 to step 5526 in which a new user-defined frame 5602 (Figure 56B) is created. Responding object 5604 is made property "responder" of new user-defined frame 5602. Property "procedure" of the method object selected in step 5524 is procedure 5606 and is made property "procedure" of new user-defined frame 5602. Property "position" of user-defined frame 5602 is initialized to be an integer 5608 whose value is zero. And property "stack" is initialized to an empty stack 5610.
Once new user-defined frame 5602 is created, processing transfers from step 5526 to step 5528 in which objects are moved from stack 5410 of user-defined frame 5304 to stack 5610 of user-defined frame 5602.
Objects are moved from stack 5410 to stack 5610 such that the order of objects is preserved. In other words, the topmost argument on stack 5410 is at the top of stack 5610, and the bottommost argument on stack 5410 is at the bottom of stack 5610. Each object is moved according to property "passage" of the corresponding constraint of property "arguments" of the retrieved operation definition.
If property "passage" is "byRef", the reference to the object is popped from stack 5410 and is pushed on to stack 5610. Thus, procedure 5606 of new user-defined frame 5602 has the same access to the argument as does procedure 5406 of requesting user-defined frame 5304.
If property "passage" is "byProtectedRef", the reference to the object is popped from stack 5410, is made a protected reference, and is pushed on to stack 5610.

WO 95/02219 ~ ~ PCT/US94/07397 Thus, procedure 5606 of new user-defined frame 5602 cannot alter the argument and the argument is therefore "readonly".
If property "passage" is "byUnprotectedRef", whether the reference to the argument on stack 5410 of requesting user-defined frame 5304 is protected is determined. Such a determination is made by querying attribute "isProtected" which is described more completely in-Appendix A. If the reference is protected, an exception of class "Reference Protected" is thrown. Otherwise, the reference is popped from stack 5410 and is pushed on to stack 5610. Thus, procedure 5606 of new user-defined frame 5602 is provided access to alter the argument passed by unprotected reference.
If property "passage" is "byCopy", a copy is made of the object found on stack 5410 and an unprotected reference to the copy is pushed on to stack 5610 of new user-defined frame 5602. Thus, procedure 5606 is given unrestricted access to a copy of the argument, which is passed by copy, but has no access to the argument itself.
Thus, new user-defined frame 5602 is complete and represents the dynamic state of the subject operation immediately preceding performance of the subject operation. Processing transfers from step 5528 (Figure 55) to step 5530 in which new user-defined frame 5602 is included in the execution state of process 5300 by being pushed on to stack 5302 (Figure 56C). Thus, stack 5410 and the method object whose property "procedure" is procedure 5406 are no longer the current stack and current method, respectively.
Stack 5610 is the current stack and the method object whose property "procedure" is procedure 5606 is the current method. Processing transfers from step 5530 (Figure 55) to step 5532. In step 5532, procedure 5606 (Figure 56C) of user-defined frame 5602 is performed.

~14~~~
As performance of procedure 5406 executes an identifier which invokes performance of the subject operation and the creation of user-defined frame 5602, performance of procedure 5606 of user-defined frame 5602 can execute identifiers which invoke further operations thereby creating further frames which are subsequently pushed on to stack 5302 and whose procedures are subsequently performed. Thus, the execution state-of process 5300 is represented by one or more frames on stack 5302, which is property "frames" of process 5300 representing a thread of execution.
After performance of procedure 5606, user-defined frame 5602 represents the dynamic state of the subject operation immediately following performance of the operation. Stack 5610 of user-defined frame 5602 contains the result, if any, produced by performance of the subject operation. Processing transfers from step 5532 (Figure 55) to step 5534 in which property "result" of the operation definition defining the subject operation is produced by querying attribute "result" of the operation definition. Processing transfers from step 5534 to result test step 5536, in which the membership of property "result" in class "Constraint" is determined. In other words, property "result" is verified to be a constraint and not a nil. If property "result" is a constraint and not a nil, the subject operation produces a result and processing transfers from result test step 5536 to a stack empty test step 5538.
In stack empty test step 5538, the engine interpreting process 5300 (Figure 56C) determines whether an object is present on stack 5610. If stack 5610 is empty, processing transfers from stack empty test step 5538 (Figure 55) to terminal step 5540. In terminal step 5540, an exception of class "Result Missing" is thrown. If stack 5610 (Figure 56C) contains an object, processing transfers from stack empty test step 5538 ~1~3~'~2 (Figure 55) to class of result test step 5542. In class of result test step 5542, the engine interpreting process 5300 (Figure 56C) determines whether the object on stack 5610 satisfies the constraint that is property "result" of the operation definition. If the object does not satisfy the constraint, processing transfers from class of result test step 5542 (Figure 55) to terminal step 5544. In terminal step 5544, an exception of class "Result Invalid" is thrown. If the object satisfies the constraint, processing transfers from class of result test step 5542 to step 5546.
In step 5546, the object is moved from stack 5610 (Figure 56C) of user-defined frame 5602 to stack 5410 of user-defined frame 5304 according to property "passage" of the operation definition as described above. Processing transfers from step 5546 (Figure 55B) to step 5548.
If, in result test step 5536, property "result" is nil, processing transfers directly to step 5548. In step 5548, new user-defined frame 5602 is popped from stack 5302 (Figure 56B) and discarded (Figure 56A). Thus, stack 5410 and the method object whose property "procedure" is procedure 5406 are again the current stack and the current method, respectively. Processing transfers from step 5548 (Figure 55B) to step 5550 in which execution of procedure 5406 is continued by incrementing integer 5408 and executing the item of procedure 5406 at the position indicated by integer 5408.
Thus, execution of an identifier which references an operation causes performance of the operation.
Operation Definition and Method Selection As discussed above, an operation definition and a method object, which together define the subject operation, are selected in steps 5510 and 5524, respectively (Figure 55A). The selection of the operation definition and method are described in the context of the WO 95!02219 PCTIUS94/07397 ~1~~~~2 embodiment of the present invention described above with respect to Figure 41B. In that embodiment, class object 4102 is formed by replicating properties "citation", "interface" and "implementation" of class definition 4100.
The selected operation definition is selected from an interface object, and the selected method object is selected from an implementation object. The operation definition and method object are each selected according to logic flow diagram 5700 (Figure 57).
The operation definition and method object, which are appropriate, depend on the class of which the responding object is an instance. The responding object is examined to determine whether the responding object is a class object, i.e., is a member of class "Class" in a class test step 5702. If the responding object is a class object, processing transfers from class test step 5702 to a class lookup step 5704. In class lookup step 5704, property "classFeatures" of the responding object and the interface superclasses of the responding object are searched as discussed below. The lexicon of class features, i.e., property "classFeatures", is searched for a key which matches the identifier of the subject operation, i.e., the executing identifier. Additionally, in class lookup step 5704, property "classMethods" of the responding object and the implementation superclasses of the responding object are searched, as discussed below, for a method object associated with the executing identifier.
Processing transfers from class lookup step 5704 to a test step 5706 in which the engine interpreting the current process, e.g., process 5300 (Figure 56A), determines whether an operation definition and a method object are found in class lookup step 5704. If an operation definition and a method object defining the subject operation are found, processing transfers from test step 5706 to terminal step 5708 where the selection process terminates. Otherwise, processing transfers from test step 5706 to instance lookup step 5710. Additionally, if the responding object is not a class object, i.e., is not a member of class "Class", processing transfers directly from class test step 5702 to instance lookup step 5710.
In instance lookup step 5710, property "instanceFeatures" of the interface object of the class of which the responding object is an instance and of the interface objects of the interface superclasses of that class are searched for a feature definition associated with the executing identifier. Also in instance lookup step 5710, property "instanceMethods" of the implementation of the class of which the responding object is an instance and of the implementations of the implementation superclasses of that class are searched for a method object associated with the executing identifier. The order in which superclasses are searched is discussed below.
It should be noted that if the responding object is a class object, a class method can only implement a class feature and an instance method can only implement an instance feature.
Processing transfers from instance lookup step 5710 to a second test step 5712 in which the engine determines whether an operation definition and a method object are found in instance lookup step 5710. If an operation definition and method object identified by the executing identifier are found, processing transfers from second test step 5712 to terminal step 5716. In terminal step 5716, the selection process terminates successfully.
Otherwise, processing transfers from second test step 5712 to terminal step 5714. In terminal step 5714, an exception of class "Feature Unavailable" is thrown.
Portions of the predefined hierarchy, which is disclosed in Appendix A, are shown in 177a Figures 58A and 58B to illustrate the order in which various classes are searched for appropriate operation definitions and method objects. While Figure 58A and 58B
show portions of the predefined interface hierarchy, the method by which the implementation hierarchy is searched is as described below with respect to the searching of the interface hierarchy.
Hierarchy graph 5800 (Figure 58A) shows class-object 5802 representing class "List". Class "List"
inherits from mix-in superclass "Ordered" and flavor superclass "Collection" represented by class objects 5804 and 5806, respectively. Class "Collection" inherits from superclass "Object" represented by class object 5808.
Class "Object" inherits from mix-in superclass "Referenced" represented by class object 5810. In Figures 58A and 58B, (i) double lines are drawn between flavor classes and subclasses of the flavor classes, (ii) single solid lines are drawn between mix-in classes and subclasses of the mix-in classes, and (iii) a dashed line is drawn between a class and an instance of the class. As discussed above, the last item of property "superclasses"
of either an interface object or an implementation object must not identify a mix-in class. The reason for this is discussed below.
Suppose for illustration purposes that the responding object in the context of logic flow diagram 5700 (Figure 57) is class object 5802 (Figure 58A) representing class "List". In class lookup step 5704 (Figure 57), property "classFeatures" of the interface object, which is property "interface" of class object 5802, is searched for an association whose key is the executing identifier. If such an association is found, the feature definition associated with the identifier is an operation definition which defines the interface of the subject operation and is selected and the search for an operation definition concludes.

WO 95/02219 Z ~ ~ PCTlUS94/07397 If, however, no such association is found, the interface object of the first superclass of class "List", namely, class "Ordered" is searched in the same manner for an appropriate operation definition. If no appropriately identified operation definition is found, the superclasses of class "Ordered" are searched in a depth-first walk of the interface hierarchy. Since class "Ordered" has no superclasses, the next item of property "superclasses" of the interface object of class "List", i.e., class "Collection", is searched as described above. Since the last superclass searched is not a mix-in class, every mix-in superclass is searched before the flavor superclass is searched.
Thus, in the manner described above with respect to class "List", the remainder of the class objects of Figure 58A are searched in the following order: class object 5808 representing class "Object" and class object 5810 representing class "Referenced".
The search for an appropriate method object defining the implementation of the subject operation is as described above for the search for an appropriate operation definition with two exceptions. First, property "classMethods" of the implementation object of a class is searched for an appropriately identified method object.
Second, the implementation superclasses, rather than the interface superclasses, of class "List" are searched for an appropriately identified method object.
The importance of the single flavor superclass being the last superclass searched is apparent if one considers the following hypothetical example. Suppose, for example, that class object 5806, which represents flavor "Collection", and class objects representing superclasses of class "Collection" are searched before searching class object 5804, which represents mix-in class "Ordered".
Suppose further than class object 5808, which represents class "Object", defines one or more features which are ..
WO 95/~9~ ~ ~ ~ ~ ~, PCT/US94/07397 adapted by class "Ordered", i.e., whose methods are supplied by class object 5804. Since class object 5808, in this hypothetical example, will always be searched before class object 5804, the method for any feature defined by class object 5808 will always be found in class object 5808. It would therefore be impossible for class object 5804 to adapt any feature defined by class object 5808. Therefore, the flavor superclass of a flavor is always the last superclass of that flavor that is searched for an operation definition or method.
Continuing in the illustrative example in which the responding object in the context of logic flow diagram 5700 (Figure 57) is class object 5802 (Figures.58A and 58B) representing class "List", hierarchy graph 5850 (Figure 58B) shows the classes searched in instance lookup step 5710 (Figure 57). The search for an operation definition and method object for an instance feature of class object 5802 (Figure 58B) differs from the search for an operation definition and method object for a class feature as described above in two ways. First, property "instanceFeatures" of the interface object of a class is searched for an appropriately identified operation definition, and property "instanceMethods" of the implementation object of a class is searched for an appropriately identified method object. Second, the classes of which the responding object, i.e., class object 5802, is a member are searched.
The manner in which the classes are searched is equivalent to that described above and is not repeated here. The following classes are searched in the following order until an appropriate operation definition and method object are found or until all classes which follow have been searched: (i) class object 5812 representing class "Class" of which class object 5802 is an instance; (ii) class object 5816 representing class "Interchanged", a mix-in superclass of class "Class"; (iii) class object 5818 representing class "Unchanged", a mix-in superclass of mix-in class "Interchanged" and therefore of class "Class"; (iv) class object 5814 representing class "Cited", a mix-in superclass of class "Class"; (v) class object 5808 representing class "Object", a flavor superclass of class "Class"; and (vi) class object 5810 representing class "Referenced", a mix-in superclass of class "Object". -If no operation definition or no method object associated with the executing identifier is found in searching the classes of Figures 58A and 58B, the operation is not defined for the responding object and an exception of class "Feature Undefined" is thrown.
As described above, if the responding object is not a class, class lookup step 5704 (Figure 57) is skipped. For example, if the responding object is list 5802A which is an instance of class "List", the classes shown in hierarchy graph 5800 (Figure 58A) are searched in instance lookup step 5712 (Figure 57) as described above.
In one embodiment of the present invention, the interface and implementation hierarchies are independent.
In other words, an interface superclass of a class is not necessarily an implementation superclass of the class, and an implementation superclass of a class is not necessarily an interface superclass of the class. An engine ensures that no feature definition is without an implementation, in such an embodiment, during performance of operation "makeClasses". Operation "makeClasses" fails, throwing an exception of class "Class Exception", upon the following condition: the class defined is not abstract, and the class defined does not define, or inherit from implementation superclasses, a method which implements a defined feature. Thus, a feature defined in the interface hierarchy is always implemented by a method found in the implementation hierarchy.

~~43~'~2 Thus, an appropriate operation definition and method object for the subject operation are found by searching the class hierarchy as discussed above or an exception is thrown.
Escalation As discussed above, a class which inherits a feature from a superclass can supersede the implementation-of the feature by providing a new implementation for the feature.
However, the new implementation may need to use the implementation provided by the superclass. An object is directed to perform a feature as implemented by a particular one or by any one of the immediate superclasses of the current class by escalating the feature. As described more completely in Appendix A, the current class is the class which provides the currently executing method.
A feature is escalated by executing a qualified identifier which references the feature to be escalated or by executing an "escalate" selector. A qualified identifier includes property "qualifier" which identifies the class which is to provide the implementation of the feature. If a feature is escalated by execution of an "escalate" selector, any immediate implementation superclass can provide the implementation of the feature.
It should be noted that no attribute is defined which provides access to property "qualifier" of a qualified identifier. Property "qualifier" of a qualified identifier is set when the qualified identifier is created and remains unaltered until the qualified identifier is destroyed. Escalation of a feature is discussed further in Appendix A.
Escalation plays a particularly important role in initializing objects. Appendix A includes a description of an initialization for each predefined class of the present invention. The initialization of an object is WO 95/02219 ~ ~~~ ~ PCTlUS94/07397 shown by logic flow diagram 5900 (Figure 59) which is discussed in the illustrative context of the initialization of a list, i.e. an instance of class "List". Operation "initialize" is escalated for each immediate implementation superclass of class "List". As discussed above, hierarchy diagram 5800 (Figure 58A) represents the interface superclasses of class "List".
Hierarchy diagram 5800 also accurately represents the implementation superclasses of class "List".
Operation "initialize", as defined by the class of which the new list is an instance, i.e. class "List", is performed in step 5902. Performance of operation "initialize" consumes zero or more arguments from the current stack. Processing transfers from step 5902 to step 5904. In step 5904, property "superclasses" of the implementation object of class "List" is produced by querying attribute "superclasses" of the implementation object. As described above, property "superclasses" of an implementation object is a list of identifiers which reference classes. Processing transfers from step 5904 to for each superclass step 5906.
For each superclass step 5906 and next step 5910 define a loop in which each superclass of property "superclasses" produced above is processed in an escalate step 5908. Thus, for each superclass, processing transfers from for each superclass step 5906 to escalate step 5908. In escalate step 5908, the superclass is directed to perform operation "initialize". according to logic flow diagram 5900. Thus, logic flow diagram 5900 is performed recursively in a depth-first walk of the implementation hierarchy. For example, the first immediate superclass, to which "initialize" is escalated in escalate step 5908, (i) performs "initialize" in step 5902, (ii) produces immediate implementation superclasses of the superclass in step 5904 and (iii) escalates "initialize" to each superclass of the superclass in the loop defined by for each superclass step 5906 and next step 5910.
Processing transfers from escalate step 5908 through next step 5910 to for each superclass step 5906.
Thus, initialization is escalated to each immediate implementation superclass in the order found in the list that is property "superclasses". Each performance of operation "initialize" by a superclass of class "List"
consumes zero or more arguments from the current stack and escalates the initialization to the immediate implementation superclasses of the superclass.
Once each immediate implementation superclass of class "List" has performed operation "initialize" according to the loop of for each superclass step 5906 and next step 5910, processing transfers from for each superclass step 5906 to terminal step 5912. In terminal step 5912, performance of operation "initialize" as defined by class "List" completes successfully. Thus, each class of which the new list is a member performs operation "initialize" in the order in which classes are searched in instance lookup step 5710 (Figure 57) described above.
Execution of Modifiers The execution of an identifier is discussed at length above and in Appendix A. However, the presence of a modifier or the nature of the identifier can alter the execution of the identifier. For example, according to the modifier substitution rules discussed in Appendix A, if the identifier identifies an attribute, operation "getAttribute"
is performed consuming as the sole argument the executing identifier. The various modifiers and the effects on the execution of an identifier are discussed in Appendix A.
184a _ ~ ~ PCT/US94/07397 Execution of Selectors A selector is a primitive whose execution causes a special action. While each selector is discussed in Appendix A, several selectors are discussed herein. For example, selectors "break", "continue" and "succeed" are discussed below in the context of performing procedures.
Execution of selector "self" pushes on to the current stack a reference to the current object, i.e. the responder of the current method. In the context of Figure 56C, the current object is object 5604 which is property "responder" of frame 5602, the current stack is stack 5610 which is property "stack" of frame 5602, and the current method is the method object whose property "procedure" is procedure 5606. Execution of selector "process" pushes on to the current stack a reference to the current process. In Figure 56C, the current process is process 5300. If the current process is an engine place, a nil is pushed on to the current stack.
Thus, an engine created according to the present invention executes procedures by sequential execution of the items of the procedures. Execution of an item of a procedure when that item is an identifier causes execution of the feature referenced by the identifier under the circumstances described above. The dynamic state of a procedure being performed is recorded in a frame. Dynamic states of methods implementing features which are user-defined are recorded in user-defined frames. A
feature definition and a method object defining the interface and implementation, respectively, of the feature are selected from the classes of which the responding object is a member. Implementations provided by superclasses of the current class are performed by escalating the feature. For example, operation "initialize" is recursively escalated such that the method provided by each class of which the responder is a member is performed. Modifiers are used to alter the execution of an identifier. And selectors provide a simple means by which commonly performed complex actions are caused by execution of a single object. These features combine to provide a great degree of functionality and generality in the disclosed set of computer instructions.
Control Constructs As discussed above, an agent can apply sophisticated logic to determine to which place or places to travel and with which agent or agents to meet. The disclosed instruction set provides a number of instructions which provide a large degree of generality enabling the realization of complex and sophisticated logic in procedures constructed of the disclosed instruction set.
The following control constructs which provide such a large degree of generality are discussed: (i) resources and resource allocation features provide processes with the ability to obtain and maintain shared or exclusive control of a resource during performance of a procedure;
(ii) decision handling features enable processes to conditionally perform procedures; (iii) exception handling features enable processes to detect and take action upon the throwing of an exception; and (iv) looping features enable a procedure to repeatedly-perform a procedure to thereby efficiently perform repetitive tasks.
Resource and Resource Allocation Features As discussed above and in Appendix A, an engine processing one or more processes does so concurrently.
For example, engine 1328 (Figure 15E) concurrently processes agents 150A and 1508. At any point during performance of a first procedure by agent 150A carried out by engine 1328, engine 1328 can temporarily suspend performance of the first procedure by agent 150A. Engine 1328 is then free to carry out performance of a second procedure on behalf of agent 1508. At any point during _ WO 95/02219 ~ PCT/US94/07397 performance of the second procedure, engine 132B can temporarily suspend performance of the second procedure on behalf of agent 150B and resume performance of the first procedure on behalf of agent 150A. In this manner, engine 132B concurrently processes agents 150A and 150B.
As discussed in Appendix A, an engine carries out performance of most of the predefined features atomically.
A feature is performed atomically if the engine refuses to suspend performance of the feature for the purpose of processing another process until the feature has completed successfully or failed. When a feature is performed atomically, no process but the current process can alter or manipulate resources since no other process will be interpreted during performance of the feature. Those predefined features which are not performed atomically are discussed in Appendix A.
There is no mechanism for causing a user-defined feature to be performed atomically. Thus, a mechanism is needed to enable a process to obtain shared or exclusive use of a resource during performance of a user-defined feature or a predefined feature which is not performed atomically. A user is able to configure a process to obtain shared or exclusive use of a resource by use of operation "use" which is defined by class "Resource" and is described in detail in Appendix A. The dynamic state of a performance of operation "use" is recorded by a use frame of class "Use Frame". Use frames are discussed in Appendix B.
Conditional Execution of Executed Objects An executed object is executed conditionally by performance of operation "if" or operation "either".
Performance of operation "if" by a responding executed object consumes a boolean object and performs the executed object if the value of the boolean object is "true." It should be noted that procedures are executed objects, WO 95/02219 PCT/US94/07397 _ i.e., members of mix-in class "Executed", and that the performance of a procedure is discussed above with respect to Figures 52A and 52B.
Logic flow diagram 6000 (Figure 60) shows the implementation of operation "if" as defined by mix-in class "Executed". In performing operation "if", the responding executed object pops a boolean object from the current stack in step 6002. Processing transfers from step 6002 to a boolean test step 6004, in which the boolean object popped in step 6002 is compared to "true".
If the boolean object is "true", processing transfers from boolean test step 6004 to step 6006 in which the responding executed object is performed. Processing transfers from step 6006 to exception test step 6010 in which the engine carrying out performance of operation "if" determines whether execution of the responding executed object in step 6006 threw an exception.
If no exception is thrown in step 6006, processing transfers from exception test step 6010 to terminal step 6008. Additionally, processing transfers from boolean test step 6004 directly to terminal step 6008 if the boolean object is "false". In terminal step 6008, operation "if" completes successfully. If an exception is thrown in step 6006, processing transfers from exception test step 6010 to terminal step 6012 in which any exception thrown by performance of the responding executed object in step 6006 is thrown, causing operation "if" to fail. The dynamic state of a performance of operation "if" is recorded by a predefined frame, i.e. a member of class "Predefined Frame", which is discussed above and in Appendix B.
One of two executed objects is selected and executed by use of operation "either". An executed object performs operation "either" by consuming a second executed object and a boolean object and causing execution of the responding executed object if the boolean object's value _ WO 95/02219 PCT/US94/07397 is "true" and, causing execution of the second executed object if the Boolean object's value is "false." Logic flow diagram 6100 (Figure 61) shows the implementation of operation "either" as defined by mix-in class "Executed".
The responding executed object pops from the current stack a second executed object in step 6102. Processing transfers from step 6102 to step 6104 in which the responding executed object pops from the current stack a Boolean object. Processing transfers from step 6104 to Boolean test step 6106 in which the Boolean popped in step 6104 is compared to "true". If the Boolean object is "true", processing transfers from Boolean test step 6106 to step 6108 in which the responding executed object is performed. Conversely, if the value of the responding Boolean object is "false", processing transfers from Boolean step 6106 to step 6110 in which the second executed object is performed. Processing transfers from step 6108 or step 6110 to terminal step 6112. In terminal step 6112, operation "either" terminates, throwing any exception thrown in step 6108 or step 6110. The dynamic state of an executed object performing operation "either"
is recorded by a predefined frame.
An executed object is selected from one or more executed objects and performed by performance of operation "select". Operation "select" is defined by class "Object"
and is performed by an object identifying the particular executed object to be performed. Figures 62A and 62B show the interface of the operation "select". Figure 62A shows the state of predefined frame 6200 which records the dynamic state of operation "select" as performed by object 6204. Predefined frame 6200 is part of the execution state of a process (not shown) and is shown immediately prior to performance of operation "select".
Object 6204 is the responder. Predefined frame 6200 has no property "responder" which identifies object 6204.
Instead, object 6204 is identified as the responder by its ~~~~~"l position within property "procedure" of the frame (not shown) immediately below the current frame on the stack (not shown) that is property "frames" of the current process. Object 6204 is the item of that procedure (not shown) whose position is one less than property "position"
of that frame. For example, the responder of user-defined frame 5602 (Figure 56C) is the item of procedure 5406 whose position is one less than the value of integer 5408.
Stack 6202 (Figure 62A) is the current stack. On stack 6202 are a mark, whose position is indicated by letter "M" and which delineates the arguments of the operation, and one or more pairs of objects and associated executed objects. For example, stack 6202 contains from top to immediately above the mark, object 6206, executed object 6208, object 6210, executed object 6212, nil 6214, and executed object 6216. Objects 6206, 6210 and nil 6214 are associated with respective executed objects 6208, 6212 and 6216.
The implementation of operation "select" is shown by logic flow diagram 6300 (Figure 63). In step 6302, objects 6206-7616 (Figure 62A) and the mark are popped from stack 6202. Processing transfers from step 6302 to step 6304 in which a dictionary is formed whose keys are objects 6206, 6210 and nil 6214 and whose associated values are respective executed objects 6208, 6212 and 6216. Processing transfers from step 6304 to step 6306.
In step 6306, the executed object whose associated key is equal to responding object 6204 is retrieved from the dictionary formed in step 6304. The retrieval of a value of a dictionary given a key of the dictionary is described in detail in Appendix A.
Processing transfers from step 6306 to a test step 6308. In test step 6308, the engine carrying out performance of operation "select" determines whether an executed object is successfully retrieved from the dictionary. If an executed object associated with a key __ WO 95/02219 PCT/US94107397 equal to responding object 6204 is found within the dictionary and retrieved, processing transfers to step 6314. In step 6314, the retrieved executed object is performed. Processing transfers from step 6314 to exception test step 6316 in which the engine determines whether the performance of the executed object completed successfully in step 6314.
If performance of the executed object in step 6314 fails and throws an exception, processing transfers from exception test step 6318 to terminal step 6320. In terminal step 6320, the exception is thrown, causing operation "select" to fail. If, on the other hand, the retrieved executed object is performed successfully in step 6314, processing transfers from exception test step 6318 to terminal step 6316. In terminal step 6316, operation "select" completes successfully.
In test step 6308, if no key equal to responding object 6204 is found in the dictionary formed in step 6304, processing transfers from test step 6308 to step 6310. In step 6310, the executed object contained within the dictionary whose associated key is a nil, e.g., executed object 6216 associated with nil 6214, is retrieved. Processing transfers from step 6310 to a second test step 6312 in which the engine determines whether an executed object is successfully retrieved in step 6310. If an executed object within the dictionary with an associated key that is a nil is found and successfully retrieved, processing transfers from second test step 6312 to step 6314 which performs the executed object as described above.
If, on the other hand, no executed object is retrieved in step 6310, no key equal to object 6204 and no key that is nil is found in the dictionary and processing transfers from second test step 6312 to terminal step 6316. As described above, in terminal step 6316, operation "select" completes successfully. Thus, if no ~~1 ~3 6'~ ~, executed object is associated with a key equal to the responding object or is associated with a nil, operation "select" completes successfully without performing any executed object in the dictionary.
Figure 62B shows the state of select frame 6200 immediately following performance of operation "select" by responding object 6204. Performance of operation "select"
produces no result as shown by empty stack 6202 of Figure 62B.
An executed object is executed repeatedly so long as a condition is met by performance of operation "while".
The dynamic state of a performance of operation "while" is recorded in a predefined frame 6402 (Figure 64). Executed object 6404 is the responder of operation "while" and is the responding executed object. Executed object 6404 is property "procedure" of predefined frame 6402. Executed object 6406 is property "precondition" of predefined frame 6402 and is the executed object consumed as an argument as discussed below.
Operation "while" is performed by responding executed object 6404 which consumes as an argument executed object 6406. Performance of operation "while" is illustrated by logic flow diagram 6500 (Figure 65). In step 6502, executed object 6406 is popped from the current stack. Processing transfers from step 6502 to step 6504, in which executed object 6406 is performed. Performance of executed object 6406 in step 6504 pushes on to the current stack a boolean which indicates whether responding executed object 6404 is to be performed.
Processing transfers from step 6504 to exception test step 6506 in which the engine carrying out performance of operation "while" determines whether performance of executed object 6406 throws an exception. If performance of executed object 6406 throws an exception, processing transfers from exception test step 6506 to terminal step 6508 in which the exception is thrown by operation _ WO 95/02219 PCT/US94/07397 "while" causing operation "while" to fail. If, on the other hand, executed object 6406 is performed successfully, processing transfers from exception test step 6506 to step 6510.
In step 6510, a boolean is popped from the current stack. Processing transfers from step 6510 to true test step 6512. In true test step 6512, the boolean popped from the current stack is compared to "true". If the boolean is "true", processing transfers from true test step 6512 to step 6514 in which responding executed object 6406 is performed. Processing transfers from step 6514 to a second exception test 6516 in which the engine determines whether performance of executed object 6404 throws an exception.
If performance of executed object 6404 in step 6514 throws an exception, processing transfers from second exception test step 6516 to a continue test step 6522. In continue test step 6522, the thrown exception is compared to an internal exception "continue", which is thrown by execution of selector "continue" as described below and in Appendix A. As described below in greater detail, execution of selector "continue" terminates an iterative performance of a procedure and initiates a new iterative' performance of the procedure. If the exception thrown in step 6514 is internal exception "continue", processing transfers from continue test step 6522 to step 6504, which is described above.
Conversely, if the exception thrown in step 6514 is not internal exception "continue", processing transfers from continue test step 6522 to a break test step 6524.
In break test step 6524, the exception thrown in step 6514 is compared to an internal exception "break", which is thrown by execution of selector "break" as described below and in Appendix A. As described below in greater detail, execution of selector "break" terminates an iterative performance of a procedure without initiating a new ~143~~~
iterative performance of the procedure. If the exception thrown in step 6514 is internal exception "break", processing transfers from continue test step 6522 to terminal step 6520, in which operation "while" completes successfully as described below.
Conversely, if the exception thrown in step 6514 is not internal exception "break", processing transfers from break test step 6524 to terminal step 6518. In terminal step 6518, the exception thrown in step 6514 is thrown causing operation "while" to fail.
If, on the other hand, performance of executed object 6404 succeeds and no exception is thrown in step 6514, processing transfers from second exception test step 6516 to step 6504 in which executed object 6406 is again performed.
Performance of steps 6504, 6506, 6510, 6512, 6514, 6516, and 6522 is repeated until performance of executed object 6406 produces a boolean whose value is "false" or until an exception is thrown in step 6508 or step 6518.
If performance of executed object 6406 produces a boolean that is "false", processing transfers from true test step 6512 to terminal step 6520 in which operation "while"
completes successfully. Thus, procedure 6406 is repeatedly performed while performance of executed object 6406 produces a "true" boolean.
Thus, operations "if", "either", "select" and "while"
provide the disclosed set of computer instructions substantial decision making capabilities.
Exception Handling Features As discussed in greater detail in Appendix A, an executed object which performs operation "do" is thereby performed, and if performance of the executed object throws an exception, the exception is thrown by operation "do", causing operation "do" to fail. The propagation of exceptions is illustrated in steps 5210 and 5212 of logic WO 95/02219 ~ PCT/US94107397 flow diagram 5200 (Figure 52A) as discussed above.
Failure of a performance of an executed object generally causes the failure of any executed object invoking, either directly or indirectly, the performance. Any exception which causes operation "live" of a process to fail causes destruction of the process. It is therefore preferred that instructions be provided which detect and prevent the propagation of exceptions.
A first executed object, e.g.', a procedure, invoking the performance of a second executed object is capable of preventing failure of the first executed object due to an exception thrown by the second executed object by "catching" any exception thrown by the second executed object. To "catch" an exception is to detect the throwing of the exception and to direct that specific action be taken in such an event. An exception which is thrown by the second executed object and which is caught by the first executed object does not cause the first executed object to fail.
An exception thrown by performance of an executed object is caught by causing performance of the executed object by use of operation "catch" in the place of operation "do". An executed object, whose performance is caused by operation "catch", is performed as if the executed object were performing operation "do" except that certain exceptions thrown by performance of the executed object do not cause operation "catch" to fail. In such a case, the exception is pushed on to the current stack and returned as a result.
The interface of operation "catch" is shown by Figures 66A and 66B. Figure 66A shows the state of catch frame 6602, which records the dynamic state of operation "catch" as defined by mix-in class "Executed", immediately prior to performance of operation "catch" by executed object 6604. Catch frame 6602 is part of the execution state of a process (not shown). Executed object 6604 is WO 95/02219 PCT/fJS94107397 the responding executed object and is property "procedure"
of catch frame 6602. Performance of operation "catch"
consumes a single argument, class object 6608 which is at the top of stack 6606. Stack 6606 is the current stack.
Stack 6606 is property "stack" of the topmost user-defined frame (not shown) below catch frame 6602 on the stack (not shown) that is property "frames" of the current process.
Class object 6608 represents class "Exception" or a subclass thereof.
The implementation of the operation "catch" is shown by logic flow diagram 6700 (Figure 67). Class object 6608 is popped from stack 6606 and thereby consumed in step 6702. Processing transfers from step 6702 to step 6704 in which executed object 6604 is performed.
Processing transfers from step 6704 to exception test step 6706 in which the engine carrying out performance of operation "catch" determines whether performance of executed object 6604 throws an exception. If performance of executed object 6604 does not throw an exception, processing transfers from exception test step 6706 to step 6708. In step 6708, a nil is pushed on to stack 6606. Processing transfers from step 6708 to terminal step 6710 in which operation "catch" completes successfully.
If execution of executed object 6604 in step 6704 fails and throws an exception, processing transfers from exception test step 6706 to member test step 6712. In member test step 6712, the engine checks membership of the exception thrown in step 6704 in the class represented by class object 6608. If the thrown exception is not a member of the class represented by class object 6608, processing transfers from member test step 6712 to terminal step 6714. In terminal step 6714, the exception thrown in step 6704 is thrown by operation "catch" and operation "catch" fails.

_ WO 95/02219 ~ ~ ~ ~ PCT/US94/07397 If the thrown exception is a member of the class represented by class object 6608, processing transfers from member test step 6712 to step 6716. In step 6716, the exception thrown in step 6704 is pushed on to stack 6606 and processing transfers from step 6716 to terminal step 6710. As described above, in terminal step 6710, operation "catch" completes successfully.
Thus, the thrown exception is caught, and operation "catch" succeeds and does not propagate the thrown exception.
Figure 66B shows the state of catch frame 6602 immediately following performance of operation "catch" by procedure 6604. At the top of stack 6606 is exception 6610 which is the exception thrown by performance of procedure 6604 if performance of procedure 6604 failed. In such a case, exception 6610 is a member of the class represented by class object 6608 (Figure 66A). In place of exception 6610 (Figure 66B) on stack 6606 is a nil (not shown) if performance of procedure 6604 succeeded. Thus, the disclosed set of computer instructions is provided with means for detecting and handling exceptions.
Looping Features The disclosed computer instruction set provides means for repeatedly performing an executed object, e.g., a procedure, facilitating performance of repetitive tasks.
Operation "while" discussed above is one example of a looping feature. Additionally, an executed object is performed repeatedly using operations "loop" and "repeat".
Operation "loop" performs an executed object indefinitely by repeatedly performing the responding executed object. Performance of operation "loop" by an executed object consumes no arguments and produces no result. Performance of operation "repeat" by an executed object consumes an integer argument and produces no ~1~36'~~
result, performing the responding executed object a number of times equal to the value of the integer argument consumed. Each performance of the responding executed object in the course of performing operation "loop" or operation "repeat" is herein called an "iterative performance" of the executed object.
Logic flow diagram 6800 (Figure 68) shows the implementation of operation "loop" as performed by-the responding executed object. The responding executed object is performed in step 6802. Processing transfers from step 6802 to exception test step 6804 in which the engine carrying out performance of operation "loop"
determines whether performance of the responding executed object throws an exception. If no exception is thrown, i.e., if the responding executed object is performed successfully, processing transfers from exception test step 6804 to step 6802 in which the responding executed object is performed again. Otherwise, if performance of the responding executed object throws an exception, processing transfers from exception test step 6804 to continue test step 6806. In continue test step 6806, the exception thrown by performance of the responding executed object in step 6802 is compared to internal exception "continue".
Internal exceptions "continue" and "break" are thrown by execution of selectors "continue" and "break", respectively, which are discussed more completely below and in Appendix A. These exceptions are called "internal"
because internal exceptions "continue" and "break" are detected by an engine which takes action in response thereto. A third internal exception, i.e., internal exception "succeed" which is thrown as a result of executing selector "succeed", is discussed below.
Internal exceptions are not caught by operation "catch" -and do not cause operation "live" to fail, thereby causing destruction of a process, except as described below.

WO 95/02219 ~ 1 PCT/US94/07397 If the exception thrown by performing the executed object in step 6802 is internal exception "continue", processing transfers from continue test step 6806 to step 6802 in which the responding executed object is performed again. Otherwise, if the exception thrown in not internal exception "continue", processing transfers from continue test step 6806 to break test step 6808. In break test step 6808, the exception thrown by performing the responding executed object is compared to the internal exception "break". If the exception thrown by performance of the executed object in step 6802 is internal exception "break", processing transfers from break test step 6808 to terminal step 6810 in which operation "loop" completes successfully. Otherwise, if the exception thrown is not internal exception "break", processing transfers from break test step 6808 to terminal step 6812 in which operation "loop" fails and the exception thrown by performance of the responding executed object in step 6802 is thrown by operation "loop".
Figure 69 shows a repeat frame 6902 which records the dynamic state of a performance of operation "repeat" by executed object 6904. Repeat frame 6902 is a part of the execution state of a process (not shown). Repeat frame 6902 includes properties "procedure", "repetitions", "repetitionsSoFar" and "position". Executed object 6904 is the responder of operation "repeat" and is property "procedure" of repeat frame 6902. Integers 6910 and 6911 are properties "repetitionsSoFar" and "repetitions", respectively, of repeat frame 6902. Integer 6910 represents the number of completed iterative performances of executed object 6906, and integer 6911 represents the total number of iterative performances specified by the consumed integer argument.
Logic flow diagram 7000 (Figure 70) shows the implementation of operation "repeat" as performed by executed object 6904. In step 7002, integer 6910 is initialized to zero and integer 6911 is initialized to the value of the integer consumed as an argument. Processing transfers from step 7002 to test step 7004 in which integer 6910 is compared to integer 6911. If integer 6910 is greater than or equal to integer 6911, processing transfers from test step 7004 to terminal step 7018 in which operation "repeat" completes successfully.
Therefore, if the integer consumed is non-positive;
performance of the operation "repeat" has no effect.
Otherwise, if integer 6910 is less than integer 6911, processing transfers from test step 7004 to step 7005 in which integer 6910 is incremented. Processing transfers from step 7005 to step 7006 in which an integer object whose value is that of integer 6910 is pushed on to the current stack. Processing transfers from step 7006 to step 7008 in which responding executed object 6904 is performed. It is up to executed object 6904 to pop the integer from the current stack. In other words, if performance of executed object 6904 does not pop the integer from the current stack, the integer remains on the current stack during performance of operation "repeat".
Processing transfers from step 7008 to an exception test step 7010 in which the engine carrying out performance of operation "repeat" determines whether performance of responding executed object 6904 throws an exception. If no exception is thrown, processing transfers from exception test step 7010 to test step 7004.
Conversely, if an exception is thrown, processing transfers from exception test step 7010 to continue test step 7014 in which the exception is compared to internal exception "continue". If the exception thrown is internal exception "continue", processing transfers from continue test step 7014 to test step 7004. Thus, if no exception is thrown in step 7008 or if the exception thrown is internal exception "continue", processing transfers to test step 7004 in which integer 6910 is again compared to - WO 95/02219 ~ ~ PCTlUS94/07397 integer 6911. Thus, execution of selector "continue"
prematurely terminates an iterative performance of responding executed object 6904 without affecting subsequent iterative performances in performance of operation "repeat".
If, in continue test step 7014, the engine determines that performance of responding executed object 6904 in step 7008 throws an exception other than internal exception "continue", processing transfers from continue test step 7014 to break test step 7016. In break test step 7016, the exception thrown is compared to internal exception "break". If the exception thrown is internal exception "break", processing transfers from break test step 7016 to terminal step 7018. In terminal step 7018, operation "repeat" completes successfully. Thus, execution of selector "break" terminates an iterative performance of responding executed object 6904 and successfully terminates performance of operation "repeat", aborting any remaining subsequent iterative performances in performance of operation "repeat".
If the exception thrown is not internal exception "break", processing transfers from break test step 7016 to terminal step 7020. In terminal step 7020, the exception thrown by performance of responding executed object 6904 is thrown by operation "repeat" causing operation "repeat"
to fail and to propagate the exception. As discussed above, operation "catch" can be used to prevent the propagation of the exception.
Once integer 6910 is decremented in step 7005 to a value less than or equal to zero, processing transfers from test step 7004 to terminal step 7018 in which operation "repeat" completes successfully. Performance of operation "repeat" by an executed object produces no results other than those resulting from performance of the executed object in step 7008 of Figure 70.

Execution of selector "continue" throws internal exception "continue" which is caught by operations "while", e.g., in continue test step 6522 (Figure 65);
"loop", e.g., in continue test step 6806 (Figure 68); and "repeat", e.g., in continue test step 7014 (Figure 70).
The throwing of internal exception "continue" by execution of selector "continue" terminates an iterative performance of an executed object performing operation "while", "loop"
or "repeat". As described more completely above, detection of internal exception "continue" during performance of operation "while", "loop", or "repeat"
causes a subsequent iterative performance of the responding procedure, e.g., in continue test steps 6522, 6806, and 7014 of Figures 65, 68, and 70, respectively.
Execution of selector "break" throws~internal exception "break" which is caught by operations "while", e.g., in break test step 6524 (Figure 65); "loop", e.g., in break test step 6808 (Figure 68); and "repeat", e.g., in break test step 7016 (Figure 70). As with the throwing of internal exception "continue" discussed above, the throwing of internal exception "break" terminates an iterative performance of an executed object in the course of performing operation "while", "loop", or "repeat".
However, throwing internal exception "break" terminates performance of operation "while", "loop", or "repeat"
rather than causing a subsequent iterative performance of the responding executed object.
Internal exceptions "continue" and "break" are distinguishable from other exceptions only in the context of operations "while", "loop", and "repeat". Therefore, in the context the following discussion of internal exceptions "continue" and "break", operations "while", "loop", and "repeat" are collectively referred to herein as "the relevant operations". As each of the relevant operations detects and takes action in response to internal exceptions "continue" and "break", internal 2~~~~~'~2 exceptions "continue" and "break" are not thrown by either of the relevant operations as are other exceptions.
Figure 71 serves as an illustrative example.
Stack 7104 is a property of process 7102 and represents the thread of execution of process 7102. User-defined frame 7106 is at the bottom of stack 7104 and . records the dynamic state of a performance of operation "live". The remaining contents of stack 7104 from bottom to top are repeat frame 7108, predefined frame 7110, which records the dynamic state of a performance of operation "loop", and user-defined frame 7112. Thus, operation "live" in Figure 71 caused performance of operation "repeat" which caused performance of operation "loop"
which caused performance of yet another feature.
Suppose that performance of the feature, whose dynamic state is recorded by user-defined frame 7112, executes selector "continue" or selector "break". As discussed above, execution of either selector "continue"
or "break" throws a corresponding internal exception. Any operation which is not designed to catch these internal exceptions, i.e., any operation other than either of the relevant operations, behaves as if an ordinary exception has been thrown (see discussion above regarding exception test step 5210 and terminal step 5212 of Figure 52A). In this situation, the internal exception is propagated by performance of the feature represented by user-defined frame 7112 and is caught by performance of operation "loop" represented by predefined frame 7110 in the manner discussed above. The internal exception is not propagated by operation "loop" represented by loop frame 7110 and therefore has no effect on the performance of operation "repeat" represented by repeat frame 7108.
If no relevant operation is currently being performed, i.e., is not present on stack 7104, execution of either selector "continue" or selector "break" produces an exception of class "Loop Missing". An exception of class "Loop Missing" is not an internal exception and is therefore propagated like any other exception.
If internal exception "succeed" is thrown, i.e., if selector "succeed" is executed, performance of the current method terminates successfully. It should be noted that, since predefined features are implemented directly, i.e., in the computer instructions which collectively form an engine, only user-defined features are implemented by method objects. Therefore, if operation "loop", the dynamic state of a performance of which is recorded in predefined frame 7110 (Figure 71), executes selector "succeed", an internal exception "succeed" is thrown, terminating performance of operation "loop". Since operation "loop" is predefined, operation "loop" does not catch internal exception "succeed" and therefore propagates internal exception "succeed", causing the termination of operation "repeat", the dynamic state of a performance of which is recorded in repeat frame 7108.
Similarly, as operation "repeat" is predefined, operation "repeat" does not catch internal exception "succeed", causing the termination of operation "live", the dynamic state of a performance of which is recorded in user-defined frame 7106. However, since operation "live" has no predefined implementation, operation "live" terminates successfully, i.e., does not fail by throwing an exception. Similarly, if the feature whose method is the subject of user-defined frame 7112 throws internal exception "succeed", i.e., executes selector "succeed", the feature terminates successfully and internal exception "succeed" is caught. In such a case, internal exception "succeed" has no effect on predefined frame 7110, repeat frame 7108, and user-defined frame 7106.
The operations discussed above, i.e., operations "do", "if", "either", "while", "select", "catch" "loo "
P
and "repeat" are performed in the context of the current method. As discussed above, predefined operations are not _ WO 95/02219 ~ ~ 4 3 ~ ,~ 2 PCT/US94/07397 implemented by method objects. The above-listed operations are all predefined and are therefore not implemented by method objects. The performance of the above-listed operations in the context of the current method is easily described by way of example.
Suppose, for example, that the feature whose method is the subject of frame 7112 (Figure 71) requests performance of one of the above-listed predefined operations, the performance of which is recorded in a predefined frame (not shown) immediately above user-defined frame 7112 on stack 7104. During performance of the predefined operation, property "stack" of user-defined frame 7112 remains "the current stack." Thus, performance of the predefined operation pops arguments from, and pushes results on to, property "stack" of user-defined frame 7112. Therefore, the predefined operations listed above are performed in the context of the current method.
The operations discussed above, i.e., operations "do", "if", "either", "while", "select", "catch" "loo "
P , "repeat", "continue" and "break", provide the object-oriented instruction set of the present invention with a substantial degree of generality and versatility.
Thus, agents are capable of employing sophisticated logic in traveling from one place to another, depositing and gathering information by interacting with agents found at various places.
Thus, a novel set of computer processes are provided which can direct their own movement through a computer network, can send multiple active copies of themselves throughout the network and can interact with other computer processes found at remote network nodes to thereby transfer information between processes. The disclosed set of computer instructions of which the novel set of computer processes are formed is portable and general. The set of computer instructions is implemented uniformly on each node of the computer network. Classes 21436'72 are made objects within the disclosed instruction set.
The instructions of a process in the present invention are interpreted rather than compiled. Control constructs are provided for resource management, decision handling, exception handling, and loop constructs. Thus, a portable and general set of computer instructions is provided from which the novel set of computer processes are constructed.
It should be understood that, while a particular set of computer instructions is described herein and in the Appendices, the present invention is not limited to the functionality of the instructions described. Therefore, the scope of the present invention shall be limited only by the claims which follow.

WO 95/02219 ~3~
~ ~ .~ ~ ~ '~ 2 PCTIUS94/07397 APPENDIX A
Copyright ~ General Magic, Inc. 1991, 1992, 1993. All rights reserved.
The following table of contents is to assist the reader in understanding the organization of and locating information within this appendix.
Table of Contents 1 Introduction 1.1 Telescript Instruction Set 1.2 Telescript Engine 1.3 Telescript Network 151.4 This Appendix 1.4.1 Scope 1.4.2 Conformance 1.4.3 Conventions 1.4.4 Organization 201.4.5 Road Map 1.4.6 References 2 Telescript Concepts 2.1 Models 2.2 Object Model 252.2.1 Objects 2.2.2 References 2.2.3 Classes 2.2.4 Inheritance 2.2.5 Features 302.2.6 Attributes 2.2.7 Operations 2.2.8 Exceptions 2.2.9 Constraints 2.2.10 Properties 352.2.11 Copying 2.2.12 Object Initialization ~14~~i'~~
2.2.13 Object Finalization 2.2.14 Class Construction 2.3 Execution Model 2.3.1 Methods 2.3.2 Procedures 2.3.3 Executed Objects 2.3.4 Identifiers 2.3.5 Static Substitution Rules 2.3.6 Dynamic Substitution Rules 2.3.7 Selector Execution 2.3.8 Modifier Execution 2.3.9 Identifier Execution 2.3.10 Method Selection 2.3.11 Method Performance 2.4 Process Model 2.4.1 Processes 2.4.2 Phases 2.4.3 Threads 2.4.4 Resources 2.4.5 Permits 2.4.6 Ownership 2.4.7 Cloning 2.4.8 Branding 2.4.9 Contacts 2.4.10 Isolation 2.4.11 Termination 2:5 Network Model 2.5.1 Agents 2.5.2 Places 2.5.3 Trips 2.5.4 Tickets 2.5.5 Meetings 2.5.6 Petitions 2.5.7 Occupation 2.5.8 Contacts 2.5.9 Citations WO 95/02219 ~ PCT/US94/07397 2.5.10 Telenames 2.5.11 Teleaddresses 2.5.12 Interchange 2.6 Timekeeping Model S 2.6.1 Time 2.6.2 Calendar Time 2.7 Pattern Matching Model 2.7.1 Patterns -2.7.2 Structure 2.7.3 Other Non-terminals 2.7.4 Metacharacters 3 Telescript Class Overviews 3.1 Groups 3.2 Kernel Group 3.2.1 Class 3.2.2 Constrained 3.2.3 Constraint 3.2.4 Exception 3.2.5 Executed 3.2.6 Execution Exception 3.2.7 Identifier 3.2.8 Kernel Exception 3.2.9 Mark 3.2.10 Modifier 3.2.11 Nil 3.2.12 Object 3.2.13 Package 3.2.14 Procedure 3.2.15 Programming Exception 3.2.16 Qualified Identifier 3.2.17 Referenced 3.2.18 Selector 3.2.19 Unchanged 3.2.20 Unexpected Exception 3.2.21 Verified 3.3 Primitive Group ~143~'~~
3.3.1 Bit 3.3.2 Boolean 3.3.3 Cased 3.3.4 Character 3.3.5 Integer 3.3.6 Number 3.3.7 Octet 3.3.8 Ordered 3.3.9 Primitive 103.3.10 Primitive Exception 3.3.11 Telenumber 3.3.12 Time 3.4 Collection Group 3.4.1 Association 153.4.2 Bit String 3.4.3 Collection 3.4.4 Collection Exception 3.4.5 Constrained Dictionary 3.4.6 Constrained List 203.4.7 Constrained Set 3.4.8 Dictionary 3.4.9 Hashed 3.4.10 Lexicon 3.4.11 List 253.4.12 Octet String 3.4.13 Set 3.4.14 Stack 3.4.15 Stream 3.4.16 String 303.5 Class Definition Group 3.5.1 Attribute 3.5.2 Class Definition 3.5.3 Class Exception 3.5.4 Feature 353.5.5 Implementation 3.5.6 Interface 3.5.7 Method 3.5.8 Operation 3.6 Identification Group 3.6.1 Citation 3.6.2 Cited 3.6.3 Named 3.6.4 Teleaddress 3.6.5 Telename 3.7 Process Group 3.7.1 Contact 3.7.2 Contacted 3.7.3 Permit 3.7.4 Process Exception 3.7.5 Process 3.7.6 Resource 3.8 Agent and Place Group 3.8.1 Agent 3.8.2 Authenticator 3.8.3 Interchanged 3.8.4 Means 3.8.5 Place 3.8.6 Ticket 3.8.7 Ticket Stub 3.8.8 Trip Exception 3.8.9 Unmoved 3.8.10 Way 3.9 Meeting Group 3.9.1 Meeting Exception 3.9.2 Meeting Place 3.9.3 Petition 3.9.4 Petitioned 3.10 Miscellaneous Group 3.10.1 Calendar Time 3.10.2 Miscellaneous Exception 3.10.3 Pattern 3.10.4 Random Stream 3.10.5 Real 4 Telescript Class Details 4.1 Conventions 4.2 Agent 4.3 Association 4.4 Attribute 4.5 Authenticator 4.6 Bit 4.7 Bit String 104.8 Boolean 4.9 Calendar Time 4.10 Cased 4.11 Character 4.12 Citation 154.13 Cited 4.14 Class 4.15 Class Definition 4.16 Class Exception 4.17 Collection 204.18 Collection Exception 4.19 Constrained 4.20 Constrained Dictionary 4.21 Constrained List 4.22 Constrained Set 254.23 Constraint 4.24 Contact 4.25 Contacted 4.26 Dictionary 4.27 Exception 304.28 Executed 4.29 Execution Exception 4.30 Feature 4.31 Hashed 4.32 Identifier 354.33 Implementation 4.34 Integer WO 95/02219 ~ ~ j~ 2 PCT/US94/07397 4.35 Interchanged 4.36 Interface 4.37 Kernel Exception 4.38 Lexicon 4.39 List 4.40 Mark 4.41 Means 4.42 Meeting Exception 4.43 Meeting Place 4.44 Method 4.45 Miscellaneous Exception 4.46 Modifier 4.47 Named 4.48 Nil 4.49 Number 4.50 Object 4.51 Octet 4.52 Octet String 4.53 Operation 4.54 Ordered 4.55 Package 4.56 Pattern 4.57 Permit 4.58 Petition 4.59 Petitioned 4.60 Place 4.61 Primitive 4.62 Primitive Exception 4.63 Procedure 4.64 Process 4.65 Process Exception 4.66 Programming Exception 4.67 Qualified Identifier 4.68 Random Stream 4.69 Real 4.70 Referenced ~14~~'~2 4.71 Resource 4.72 Selector 4.73 Set 4.74 Stack 4.75 Stream 4.76 String 4.77 Teleaddress 4.78 Telename 4.79 Telenumber 104.80 Ticket 4.81 Ticket Stub 4.82 Time 4.83 Trip Exception 4.84 Unchanged 154.85 Unexpected Exception 4.86 Unmoved 4.87 Verified 4.88 Way 5 Telescript Syntax 205.1 Telescript 5.2 Character Telescript 5.2.1 Preface and Comment 5.2.2 Executed Objects 5.2.3 Other Non-terminals 255.3 Binary Telescript 5.3.1 Preface and Comment 5.3.2 Executed Objects 5.3.3 Other Non-terminals 5.4 Numeric Codes 305.4.1 Predefined Classes 5.4.2 Predefined Features 5.4.3 Executed Object Encodings 6 Syntax of Module 6.1 General Structure 356.2 Detailed Structure 6.2.1 Module _ WO 95/02219 s r' PCT/US94/07397 ~~~~~iN
6.2.2 Interface 6.2.3 Feature 6.2.4 Constraint 6.2.5 Other Non-terminals S 7 Predefined Module 8 Predefined Class Graph The three major elements of the Telescript technology are introduced in this section of this appendix: the Telescript Instruction Set, i.e. the subject of this appendix; the Telescript Engine, which implements the Instruction Set; and the Telescript Network, formed by interconnecting Engines.
1.1 Telescript Instruction Set The Telescript Instruction Set, which is sometimes referred to herein as the "Instruction Set", is a set of computer instructions which collectively form a programming language whose intended field of application is distributed systems and applications. The Instruction Set is both object-oriented and interpreted. In Microfiche Appendices E and F, the Instruction Set is sometimes referred to as "the Language".
The most distinctive classes built into the Instruction Set are classes "Agent" and "Place". An agent, i.e. a computer process that is a member of class "Agent", is an active object that can examine and modify itself, transport itself from one place in a network to another, and interact with the other agents it finds . there. This power is counterbalanced by permits, which enable either a programmer or an administrator to grant only particular capabilities to particular agents on particular occasions. A place, i.e. a computer process that is a member of class "Place", is an active object '~1~~'~'~
that can examine and modify itself and that serves as a venue for agents and a context in which they can interact.
An agent goes to a place and there interacts with the place and other occupants of the place. Agents and places cannot interact at a distance. Thus the Instruction Set implements remote programming (RP), not the more familiar remote procedure calling (RPC) paradigm. RP improves upon RPC by enabling system elements to interact without communicating, improving the performance of the interactions by reducing their latency. Equivalently, RP
enables system elements to customize one another by stationing their own agents--and thus themselves--in one another's domain.
The Instruction Set strives to achieve these characteristics:
Safety The Instruction Set prevents a process from exceeding its permit, interfering with another process without the latter's permission, or directly manipulating the computer on which the process runs. This helps to avoid viruses.
Portability The Instruction Set makes no concessions to the hardware or software constraints or peculiarities of a particular computer system. This enables a process to be executed anywhere within or around a network.
Extendibility The Instruction Set gives to types of information object defined by the programmer the same stature as those built into the Instruction Set. This enables extension of the Instruction Set for particular purposes.
Elevation The Instruction Set draws no distinction between volatile and non-volatile storage. Every information object is inherently persistent. This increases a process level of abstraction and decrease the process' size.
1.2 Telescript Engine A program written in the Instruction Set is brought to life by a Telescript Engine, which is sometimes referred to herein as an "Engine", which interprets the instructions the program contains. Such a program is called a telescript. One Engine can interpret many telescripts concurrently.
An Engine can implement the abstractions of the Instruction Set without directly relying upon the hardware or operating system of the computer on which the Engine runs. The Engine does this using standard interfaces to the communication, storage, and other subsystems the Engine requires of the platform. These form the Telescript Communication Application Programming Interface (API) which is attached as Appendix C and is a part of this disclosure.
An Engine can implement privileged escapes from the abstractions of the Instruction Set. Such escapes enable the construction, external to the Engine, of operational, administrative, and managerial (OAM) tools.
Such tools are vital to the success of large-scale communication systems.
1.3 Telescript Network Two or more Engines can be interconnected. The resulting Telescript Network, which is herein sometimes called the "Network" is the universe within which agents travel. The Network encompasses computer systems that would be considered clients of other networks, not parts of them.

A computer system is a part of the Network only if the system incorporates an Engine and thus provides places from and to which agents can go. This requirement makes the Network homogeneous with respect to the structure, as well as the transport, of information.
Engines are interconnected so that they can move agents among themselves. Agents are serialized, or encoded, for this purpose according to the Telescript Encoding Rules described in Appendix B. The resulting octet string is transported between Engines as prescribed by the Platform Interconnect Protocol which is attached as Appendix F as a part of this disclosure.
1.4 This Appendix This appendix is described as follows.
1.4.1 Sc- ope This appendix defines Version 0.8 of the Instruction Set. The Telescript Encoding Rules, the Telescript Protocol, and the Telescript API are all beyond the scope of this appendix and are the subject of respective Appendices B, F, and C.
One may view the Instruction Set as an instruction set for a virtual machine. From this viewpoint, one can readily envision still higher-level languages, compilers for which produce telescripts. General Magic Inc. of Mountain View, California is developing such a language which is called High Telescript. In the context of Appendices A-F of this disclosure, the Instruction Set itself is sometimes called Low Telescript.

1.4.2 Conformance A manufacturer of an Engine has to satisfy certain requirements to properly claim conformance to this appendix.
1.4.3 Conventions 218a -- WO 95/02219 ~ PCT/US94/07397 These conventions of prose are followed throughout this appendix:
t "The a attribute" means "the attribute whose identifier equals a".
~ "Is a key" means "equals a key", and not "is the same as a key".
"An x" generally means "a member of X", and not !'an instance of X".
"Throws x" generally means "throws a member of X".
t "Whether statement" means "statement iff true".
Iff means "if and only if", as it does in mathematics.
MSB stands for "most significant bit".
LSB stands for "least significant bit".
Note: Throughout this appendix, notes (like this) are explanatory, not definitive.
1.4.4 Orctanization This appendix is divided into eight sections.
Section 1 is this introduction. Section 2 introduces the Instruction Set's major concepts. Section 3 overviews the Instruction Set's predefined classes. Section 4 defines them in detail. Section 5 defines the Instruction Set's syntax.
Section 6 defines the conventions followed by the formal definitions of the interfaces to the predefined classes. Section 7 comprises those formal definitions.
Section 8 shows the part of the class graph that comprises the predefined classes and no user-defined classes.
1.4.5 Road Map Parts of this appendix are definitive, while others are merely informative. The definitive sections are Sections 1, 2, 4, and 5. In contrast, Sections 3, 7 and 8, reorganize material in Section 4, while Section 6 duplicates aspects of High Telescript.

~~~J~~~
Different audiences require different parts of this appendix. Those interested only in the Instruction Set's scope and structure need read only Sections 1, 2, and 3.
A Telescript programmer requires Section 4 as well. A Low Telescript programmer must master also the portion of Section 5 covering character telescripts, while the High Telescript programmer need not do so. A High Telescript Compiler or Engine implementor requires the portion of Section 5 covering binary telescripts.
For Telescript practitioners, Sections 7 and 8 are indispensable references.
1.4.6 References This appendix relies upon these other documents:
[10646) Information technoloctv--Universal Coded Character Set UCS , ISO/IEC DIS 10646, International Organization for Standardization and International Electrotechnical Commission, 1990.
lUnicode) The Unicode Standard: Worldwide Character Encoding, Volume 1, Version 1.0, The Unicode Consortium, Addison-Wesley, 1991.

The Instruction Set defines a variety of concepts, the most important of which are introduced in this section of this appendix. These concepts are divided into "models". A subsection is devoted to each model. Within each subsection, a lesser subsection is devoted to each concept in the model.
Being for remote programming, the Instruction Set includes concepts spanning the realms of languages, operating systems, and networks. Conventionally separate, these areas, in this instruction set are brought together.
2.1 Models The Instruction Set's concepts are divided into models. This division serves pedagogic purposes alone and is not visible, in any sense, at run time.
The following models are defined:
Obj ect The "Object" model provides object orientation, e.g., objects, references, classes, operations, and exceptions.
Execution The "Execution" model provides sequential execution, e.g. methods, procedures, and identifiers.
Process The "Process" model provides multi-tasking, e.g., processes, resources, permits, contacts, and ownership.
Network The "Network" model provides network architecture, e.g., agents, places, trips, meetings, and telenames.
Timekeeping The "Timekeeping" model provides means for keeping time, e.g., times and calendar times.
Pattern Matchincr The "Pattern Matching" model provides means for pattern matching, e.g., patterns.
The models are presented below in the order in which they are listed above. The concepts within each model are presented in a logical order.

2.2 Object Model The Instruction Set realizes the "object model" this section defines.
2.2.1 Ob'e An object is the Instruction Set's unit of both information and information processing. An object is an instance of a class.
Note: An object may be anything at all, either simple, e.g., a boolean, or complex, e.g., a dictionary, either passive, e.g., a string, or active, e.g., a process.
Note: In messaging, the application for which the Instruction Set was conceived, the objects can include the messaging system's component parts, e.g., its mailboxes and distribution lists; the information objects the system transfers, e.g., messages and delivery reports; and the elements of those information objects, e.g., the fields of envelopes.
Persistence Every object is "persistent". If the Engine fails and subsequently recovers, the only effect upon the object is its temporary unavailability.
Size Every object has a "size" which is the approximate amount of persistent storage the object occupies, measured in octets.
Note: An object's size may vary from place to place.
2.2.2 References ~~~3~'~2 A "reference" is the means by which an object is denoted and accessed.
Protected vsUnprotected A reference is either "unprotected" or "protected".
An unprotected reference enables the object to be changed, while a protected reference does not. An unprotected reference can be made protected, but not conversely.
Creation There exists one or more references to every object.
Creating an object creates an initial reference to the object as well. The reference is protected iff the object is immutable. Additional references can be created as desired. Created references are protected iff the source reference is protected.
Note: Many predefined operations create new references to existing objects and return the references as results.
Comparison , All unprotected references to an object are equivalent. Changes made using an unprotected reference to an object are made to the object, and thus, in effect, to all other references to the object. All protected references to an object also are equivalent to one another.
An unprotected and a protected reference differ in two ways which come to light when a protected reference is used to request a feature of the object the reference denotes. First, if the feature would change the object, the feature fails, throwing "Reference Protected".
Second, if the feature returns a reference to--not a copy of--one of the object's properties, that reference is protected.

~1~3~7~
One can determine whether two references denote the same object.
Discarding A reference should be discarded when no longer needed. If no references to the same object remain, the object itself is destroyed.
Voiding The Engine voids references under circumstances this appendix defines. A "voided" reference no longer provides access to an object. The Engine throws "Reference Void"
rather than push a voided reference onto the stack.
2.2.3 Classes A "class" defines a set of objects, the class' "instances", all having the same interface and the same implementation. The class itself has both an interface and an implementation, which are potentially unique to the class.
Predefined vs User-defined A class is either predefined or user-defined. A
"predefined" class, built into the Instruction Set and defined in this appendix, represents a kind of object available to every Telescript programmer. A
"user-defined" class, defined by the programmer, extends the Instruction Set for specific purposes.
Concrete vs Abstract A class is either concrete or abstract. A "concrete"
class can have instances. An "abstract" class cannot, but its subclasses can and often do. A concrete class, or one of its implementation superclasses, shall implement each feature either native to or inherited by the class.

WO 95/02219 ~ ~- ~ ~ ~ r~ ~ PCT/US94/07397 Sealina A class, usually concrete, can be sealed. The Engine prevents a sealed class from having user-defined, but not predefined, immediate subclasses.
Creation A concrete class can create a new instance given initialization parameters describing the new instance.
Not every instance of every class, however, can be created in this way. The remaining ones can be created by first creating one instance in this way and then modifying the instance's attributes as necessary.
Conversion A concrete class can convert an instance of one class to an instance of itself. Not every instance of every class, however, can be converted to an instance of every other class. The definition of operation "convert", and the nature of the object conversion produces, depend upon the two classes involved.
Compatibility A class is embodied as a cited object. One class is backward compatible with another iff the former's interface can be created from the latter's by making only changes of the following kinds. First, a feature can be added. Second, if in the latter's interface an attribute is read-only, that attribute's class can be narrowed to a subclass. Third, the class of an argument of an existing operation can be widened to a superclass. Fourth, the class of the result of an existing operation can be narrowed to a subclass.
2.2.4 Inheritance Native vs Inherited WO 95/02219 ~ PCTIUS94/07397 -All classes, both predefined and user-defined, are related to one another by "inheritance". Every class has various characteristics. In general, some of a class' characteristics are "native" to the class, while others are "inherited" from other classes. Inheritance is a transitive relationship. A class inherits from a class the latter's inherited characteristics, as well as the latter's native characteristics. Extending the above terminology only slightly, an object's native characteristics are those native to the class of which the object is an instance.
The Instruction Set sufficiently constrains the inheritance relationships among classes that the classes and their relationships can be described by simple graphs.
In such a graph, the nodes represent classes, the arcs between nodes the inheritance relationship between the classes the nodes represent.
Subclass vs Superclass If one class inherits characteristics from another, either directly or indirectly, the former is a "subclass"
of the latter, and the latter is a "superclass" of the former. If one class inherits characteristics from another directly, the former is an "immediate subclass" of the latter, and the latter is an "immediate superclass" of the former.
An instance of a class is a "member" of that class and of its superclasses.
Flavor Some classes are flavors. The "flavors", each of which is either abstract or concrete, form a tree as follows. The tree's root represents class "Object". Each destination node reached by an arc emanating from any given source node represents an immediate subclass of the flavor that the source node represents. Thus the source WO 95/02219 ~ PCT/US94/07397 node represents an immediate superclass of the flavor represented by each such destination node. Each destination node reached by way of one or more arcs in succession represents a subclass of the flavor that the source node represents. Thus the source node represents a superclass of the flavor represented by each such destination node.
Note: Flavors provide single inheritance.
Mix-in All other classes are mix-ins. A "mix-in" and its zero or more superclasses--all mix-ins themselves--form a second tree as follows. The tree's root is the mix-in.
The arcs represent the inverse inheritance relationship between the mix-ins that the tree's nodes represent. A
mix-in is abstract.
Note: Mix-ins provide a limited form of multiple inheritance.
Class Graph .
The flavors and mix-ins together form a directed graph. The graph is constructed in the following two steps. First, the arcs in each mix-in tree are reoriented to represent the inheritance relationship, rather than its inverse. Second, the thusly altered mix-in trees are superposed upon the flavors tree.
Note: There is one, global class graph, achieved by identifying classes globally. In any particular place, knowledge of the graph may be incomplete.
Canonical Order A class, either a flavor or a mix-in, and its superclasses, both flavors and mix-ins, have a "canonical order" which is that of a walk of a third tree.
The tree is formed as follows. Its root is the class in question, its other nodes the class' superclasses. As WO 95/02219 ~ PCT/US94107397 in a mix-in tree, the arcs represent the inverse inheritance relationship among the classes that the tree's nodes represent.
The walk is a depth-first one in which a class' immediate superclasses are visited in their canonical order, which the class defines.
Interface vs Implementation All of the above can be said about either of three sets of class characteristics: the classes' interfaces, their implementations, or both. This appendix explains, either explicitly or implicitly, which of the two sets of characteristics is being discussed at any particular point within this appendix.
Note: More often than not, the classes' interfaces are in view.
2.2.5 Features A "feature" is an externally visible characteristic of an object. Features enable objects to interact. If one object possesses a reference to another, the former can request a feature of the latter. All instances of a class have the same features, defined by the interfaces native to or inherited by that class.
A feature "succeeds", when it is used correctly, and "fails", at other times. When the feature fails, an exception is thrown.
Attribute vs Operation Features are of two kinds, attribute and operation.
Note: Whether a particular feature is made an attribute or an operation is, in part, a matter of taste.
Sealina WO 95/02219 ~ PCT/US94/07397 A class can seal a feature, either native or inherited, thereby preventing the feature's implementation by user-defined, but not predefined, subclasses.
Implementation A class can implement any native feature or any inherited feature not sealed by a superclass. In the latter case, the class' implementation supersedes the superclasses' implementations, if there are any.
Access The Engine controls access to a feature. A feature's "access" defines the objects that can use the feature, and is one of the following:
public Any object can use the feature.
private Only the object that has the feature can use it.
svstem Only the Engine itself can use the feature, which is predefined.
2.2.6 Attributes An "attribute", an object, is any characteristic that one object, the "requester", can ask another or the same object, the "responder", to get and perhaps set. An attribute that cannot be set is "read-only".
"aGet" vs "aSet"
An attribute is exactly equivalent to either one operation, if the attribute is read-only, or two, otherwise. The first operation, denoted symbolically as "aGet", gets the attribute. It has no arguments, and its ~1~3~'~~
result is the attribute. The second operation, denoted as "aSet" replaces the attribute with a particular object.
The one argument of operation "aSet" is that object, and the operation has no result.
Manual vs Automatic A class can implement an attribute either manually, automatically, or, iff the class is abstract, not at all.
To "manually" implement an attribute is to provide methods for "aGet" and, if the attribute can be set, "aSet". To "automatically" implement an attribute is to let the Engine do so, which it does if the attribute is an instance attribute, the class is concrete, and neither the class nor its implementation superclasses implement the attribute. The Engine implements the attribute with a property of the same identifier.
The Instruction Set defines the getting and setting of attributes with the aid of "getAttribute" and "setAttribute", respectively, internal operations an argument of which is the attribute's identifier ("a").
These two operations get and set, respectively, the property associated with the attribute, if the latter is implemented automatically, or perform the user-defined methods for operations "aGet" and "aSet", otherwise, i.e., if the attribute is implemented manually. An attribute's manual or automatic implementation alike is located using the execution model's algorithm for locating the implementation of an operation.
Storage Location An attribute that is not read-only behaves by default as a "storage location". That is, operation "aGet"
returns the object most recently supplied successfully as an argument of operation "aSet". Exceptions to this default behavior shall be called out in the prose descriptions of particular attributes.

DEMANDES OU BREVETS VOLUMINEUX

COMPREND PLUS D'UN TOME.
CECI EST LE TOME r DE
NOTE: Pour les tomes additionels, veuillez contacter le Bureau canadien des brevets THAN ONE VOLUME
. THIS 1S VOLUME ~ OF
NOTE: For additional volumes please contact the Canadian Patent Office

Classifications
International ClassificationG06N5/04, G06F15/16, G06F9/44, G06F13/00, G06F9/50, G06F9/46
Cooperative ClassificationG06F9/54, G06F9/4862, G06F9/465
European ClassificationG06F9/54, G06F9/46M, G06F9/48C4P2M
Legal Events
DateCodeEventDescription
9 Jul 1999EEERExamination request
8 Jul 2008MKLALapsed