A Foundational Theory for Attribute-based Communication

In open systems, i.e. systems operating in unpredictable environmentsand with components that may join or leave atany time, behaviors can arise as side effects of intensive componentsinteraction. Finding ways to understand and designthese systems and, most of all, to model the interactions oftheir components, is a difficult but important endeavor. Totackle these issues, we present AbC, a calculus for attributebasedcommunication. An AbC system consists of a set ofparallel agents each of which is equipped with a set of attributes.Communication takes place in an implicit multicastfashion, and interactions among agents are dynamically establishedby taking into account 'connections' as determined bypredicates over the attributes of agents. First, the syntax andthe semantics of the calculus are presented, then expressivenessand effectiveness of AbC are demonstrated both in termsof modeling scenarios featuring collaboration, reconfiguration,and adaptation and of the possibility of encoding channelbasedinteractions and other interaction patterns. Behavioralequivalences for AbC are introduced for establishing formal relationshipsbetween different descriptions of the same system.A Java run-time environment has been developed to supportprogramming of the above mentioned class of systems by relyingon the communication primitives of the AbC calculus.The impact of centralized and decentralized implementationsfor the underlying communication infrastructure, that mediatesthe interaction between components, has been studied.