Safe integration of new concerns in a software architecture

Abstract

Software architectures must frequently evolve to cope with changing requirements, and this evolution often implies integrating new concerns. Unfortunately, existing architecture description languages provide little or no support for this kind of evolution. The software architect must modify the architecture manually, which risks introducing inconsistencies. In previous work, we have proposed the TranSATframework, which provides a pattern construct for describing new concerns and their integration into an existing architecture. As the interaction between the new concern and the existing architecture may be complex, it is essential that the framework ensure the coherence of the resulting architecture. In this paper, we introduce a language for specifying patterns and verifications that ensure that the concern represented by a pattern can be safely integrated into an existing architecture. The verifications comprise static verifications that check coherence properties before the architecture is modified and dynamic verifications that focus on the parts of the architecture that are affected by the pattern. As a result of these verifications, patterns can be provided as a commodity, such that a software architect can confidently apply a pattern obtained from a third-party developer