Will investigate exception handling, debugging support in Java. There's a book which talks about debugging support in the sun.debug.* libraries. An will purchase the book and get reimbursed when the funds come in.
Will read Petri Net stuff and think about extensions to Petri Nets for mobile agents. Calvin will read some papers regarding agents and Petri Nets.
Finishing design of multicast algorithm. Many papers in the multicast area will be available in the AgentOS web site.
Will do a survey on routing. Fixing problem with his set-graph algorithm.
This paper analyzes the tradeoffs between performing an agent migration vs. performing an RPC (or message communication) to complete the same task. This paper reminded us that migration may not always be efficient. When the agent code size is large compared with communication data size, migration may be unnecessary.
This paper led us to consider whether we should support system-controlled migration, i.e., the system migrates the agent based on load and communication optimality. If we are to support system-controlled migration, the agent behavior model needs to consider two scenarios:
1) Only interactions between a group of agents are specified in the model. The location of specific agents are not important. This situation arises when a group of agents is responsible for carrying out a compute-intensive task, e.g., computing PI to a million digits cooperatively.
2) Explicit migrations are specified in the agent behavior model. This situation arises when the agents need to be aware of their locations, e.g., in a mobile computing situation or a network management situation.