9619522 Computing applications in the real world are generally distributed and open to interaction with their environment. Creating an architecture for such systems requires developing mechanisms for managing complex interactions between distributed components. As requirements and the environment change, both the individual components and the pattern of interactions between them evolve. This implies that interaction mechanisms may need to be synthesized during the execution of the system according to the evolving needs of components. In current software, the implementation of policies governing interactions between components is typically represented using a variety of low-level primitives -- such as message-passing, remote procedure call, or shared memory -- and is therefore a major source of code complexity. The project is studying programming techniques for abstract and modular representation of interaction patterns. Such programming techniques will allow code for distributed applications to evolve gracefully -- for example, by reusing components when only their interaction context changes. The research is based on Actors, a formal model of concurrent computation in open distributed systems. ***
