Two Ways of Implementing Software Connections among Distributed Components

International audienceSoftware architecture has emerged in the last decade. Applications are represented as set of interconnected components. The way to realize components has reached a certain maturity in both industrial and academic approaches; it has almost the same consideration or definition in the two domains. The way to implement the interconnections between components, however, is not as well understood as implementing components. The experience of implementing the interconnections between components is dispersed since interconnection models are integrated in component models. Every component model defines its own interconnection model without basing on any reference model. This makes the realizations ad-hoc and non uniformed. We propose to make more standard the realization of the connections between components and to distinguish two different entities that differ in nature. This difference of nature implies a different way in using them and a different way in implementing them. We propose to distinguish between communication abstractions embodied in components with explicit interfaces, and communication abstractions embodied in connectors with implicit interfaces. This difference enables a better understanding of the interactions and how to implement them. The realization of the load balancing communication abstraction is used to illustrate the two entities