On implementation techniques for mobile and pervasive middleware families

Mobile and pervasive computing scenarios are challenging for middleware technology, e.g. due to resource constraints, mobility, decentralized architecture, realtime dependability, etc. Common middleware systems fail in this context. They lark flexibility and configurability. Moreover, they are not able to down-scale or to be ported to the various mobile and embedded platforms. To overcome these challenges, we propose fine-grained program family-based middleware architectures. Furthermore, we analyze several analysis, design and implementation techniques to build mobile and pervasive middleware, in particular Feature-Oriented Domain Analysis, Feature-Oriented Programming, Mixin Layers and Aspect-Oriented Programming. We analyze and discuss their pros and cons. To flesh out our considerations, we introduce and refer to several proprietary studies. Our results are positive, the degree of configurability and reusability compared to common middleware is extremely high. This is achieved without constraining maintainability and comprehensibility. The resource consumption correlates with the number of supported features.