Standard Type Soundness for Agents and Artifacts

Abstract

Formal models, core calculi, and type systems, are important tools for rigorously stating the more subtle details of a language, to characterise and study its features and the correctness properties of its programs. In this paper we present {FsimpAL} (FsimpaALlong), a formal calculus modelling the agent and artifact program abstractions provided by the simpA{} agent framework. The formalisation is largely inspired by textsc{Featherweight Java}. It is based on reduction rules applied at certain evaluation contexts, properly adapted to the concurrency nature of simpA{}. On top of this calculus we introduce a standard type system and prove its soundness, so as to guarantee that the execution of a well-typed program does not get stuck. Namely, all primitive mechanisms of agents (activity execution), artifacts (field/property access and step execution), and their interaction (observation and invocation) are guaranteed to be used in a way that is structurally compliant with the corresponding definitions: hence, there will not be run-time errors due to {FsimpAL} distinctive primitives.