interActors: A Model for Separating Complex Communication Concerns in Multiagent Computations

Multiagent computations increasingly require complex and diverse interactions. Existing approaches to separating these communication concerns offer static protocols, which cannot handle dynamically evolving numbers and sets of communicating agents, and leave complex initialization steps mixed in with functional concerns. We present interActors, a model for separating complex communication concerns of applications from their functional concerns. These first-class communications are self-driven and dynamically change in response to evolving communication needs; they can be easily created, reused and composed. A prototype implementation is presented. A case study illustrates improved programmability using interActors.