I will discuss, very briefly, the group's ideas on archetypes for reactive systems. Our primary focus is message-passing systems, and our target implementation platforms are primarily C (or C++) and Unix, or Ada. Archetypes for scientific computation and combinatorics are based on the computation, communication and proof structure of classes of programs. By contrast, the archetypes we are working on for reactive systems are closer to classes that deal with specific problems such as mutual exclusion, multiprocess synchronization, distributed checkpointing, distributed clocks, collective communication operations such as atomic broadcasts, and so on.
We plan to explore archetypes for commercial applications on the National Information Infrastructure. The hypothesis here is that commerce on the NII can be categorised into a relatively small number of groups, each of which have similar applications. If this hypothesis is correct, then we can have some hope that archetype-based design will help the large numbers of businesses, particularly small businesses, that operate part of their businesses on the net.
A specific concern for reactive systems is reuse of methods of reasoning about programs, and incorporating formal reasoning with software engineering methods such as testing. This concern is raised by the subtleties of reasoning about the correctness of reactive systems.