A Reflective Forward-Chained Inference Engine to Reason about Object-Oriented Systems

Recently, a lot of research concerning co-evolution has been done. Co-evolution is a brand new approach that tries to find a solution for the problem of the missing link between the design and the implementation of object-oriented software systems. To date this has been done by using a rule-based backward chained reasoning mechanism as a declarative meta layer on top of the implementation of an object-oriented system. However, one of the major drawbacks of this approach is the fact that a goal should be clearly specified. Such a (specification of the) goal is very often not available. In this dissertation we will show that in some specific cases a rule-based backward chained reasoning mechanism is not sufficient. Therefore, as an alternative mechanism we suggest a rule-based forward chained reasoning mechanism to implement the declarative meta layer, in order to handle these specific cases. We will validate this thesis by means of some experiments done with a forward chained prototype we built. The prototype is based on the expert system tool KAN and implemented in Squeak. The conducted experiments confirm our thesis. Especially in cases where no specific goal can be specified and in cases where the reasoning state has to be preserved, our alternative approach proved to be preferable.