Software Connectors: A Hierarchical Model

To support rapid software evolution, it is desirable to construct software systems from reusable components. In this approach, the architecture of a system is described as a collection of components along with the interactions among these components. Whereas the main system functional blocks are components, the properties of the system also strongly depend on the character of the component interactions. This fact gave birth to the “connector” concept which is an abstraction capturing the nature of these interactions. The problem tackled in this paper is that even though the notion of connectors originates in the earliest papers on software architectures [24, 18], connectors are currently far from being a typical first class entity in the contemporary component-based systems. The authors address the problem by (1) bringing an additional argument in favor of considering connectors as first class entities representing component interactions and by (2) introducing a connector model reflected at all the key stages of an application’s development: ADL specification, deployment, and implementation. By articulating the “deployment anomaly”, the paper identifies the role connectors should play when the distribution and deployment of a component-based application is considered. Based on basic connector elements (both predefined and user-defined), a connector can be (at least) partially generated from its ADL generic description and from the information on deployment of the components it connects. This way, the underlying code of a component-based application is advantageously factored into the deployment-neutral part (comprising the “real” application functionality located in components) and deployment-sensitive part (embodied in connectors) which can be regenerated anytime the deployment of the application is modified. As a proof of the concept, a case study is provided, illustrating how the connector model can be integrated into the SOFA/DCUP component model.