Separating distribution policies in a shared data space system

To more fully exploit its benefits, the principle of separation of concerns should be built into the methods of designing systems as well as into the techniques that we use for implementation. We argue that for distributed applications, separating concerns should be supported by the middleware on which the applications are executed. In particular, we argue that the middleware should provide the proper mechanisms to enforce application-specific policies for different concerns. Interweaving of policies should take place at run time such that it becomes possible to change them without the need for recompiling the application source code. To illustrate that this approach is principally feasible, we present a design for a middleware based on the shared data space paradigm. Here distribution is taken as an example of a concern that can be treated separately from other concerns and transparently to applications. We present a case study that shows how we can vary the distribution policy for an application. The case study shows a trade-off between the use of bandwidth versus the speed of message delivery.