Connecting methodologies and infrastructures in the development of agent systems

In the building of agent systems developers can be supported by both appropriate methodologies and infrastructures, which guide them in the different phases of the development and provide useful abstractions. Nevertheless, we assist to a situation in which methodologies and infrastructures are not connected each other: the products of the analysis and design phases could not always be exploited in the implementation phase in a direct way, even if sometimes CASE-tools are present to help in translating methodologies' diagrams in infrastructures' code. This leads to a "gap" between methodologies and infrastructures that is likely to produce fragmented solutions and to make the application maintenance difficult. In this paper we face this issue, proposing three directions to solve the problem. We do not want to propose a "new brand" methodology and infrastructure tightly connected, rather, we aim at reusing as much as possible what already exists, not only in abstract terms, but also in concrete "fragments" of methodologies; an appropriate meta-language that describes how a methodologies works would be useful to more easily map them onto the infrastructures, or even to "compose" a new methodologies. A further approach is based on an "intermediate" layer between methodologies and infrastructures, which provides a mapping between the involved entities.