C2AS: a system supporting distributed Web applications composed of collaborating agents

In this paper we describe the Cambridge Collaborative Agent System (C²AS). This is a prototype system designed to demonstrate the functionality and basic architecture of a framework for co-ordination between different components (or agents) of distributed `Web´ applications. Co-ordination in C²AS is achieved through the use of tuple spaces, as used in Linda. However, the access primitives used in C²AS are not those used by Linda, but the BONITA primitives, which provide asynchronous access to tuple spaces as opposed to synchronous access. Because C²AS is tuple space based it supports temporal and spatial separation of agents. The prototype system supports agents written in either Java (including applets) or C