Traditionally, extensibility and adaptability in middleware is achieved through thorough design of the problem domain. The key variability points are modeled at design time to allow plugin of new functionality at different points in the system. Unfortunately, it is historically shown that the actual adaptability needs at some point differs from the ones predicted. We argue that the seams, which are hidden in middleware with seamless design, should be accessible for pervasive computing application developers. It cannot be foreseen which aspects of a seam the domain or application need. We suggest that in addition to modeling the domain variability points the middleware should expose a model of the internal processing mechanisms of the middleware itself. The middleware should support modification of this model. Furthermore, at all places where the middleware achieves the normal seamless use of the domain, the model should be adaptable and adaptations should be absorbed into the middleware. We present model-based translucency as a middleware construction goal. The supporting arguments are given and examples of use of models in middleware are compared to the requirements for which we argue.
Proceedings of the 2nd International Workshop on Middleware for Pervasive Mobile and Embedded Computing: M-mpac'10, 2010
Main Research Area:
Second International Workshop on Middleware for Pervasive Mobile and Embedded Computing, A Workshop of Middleware 2010