First class feature abstractions for product derivation

The authors have observed that large software systems are increasingly defined in terms of the features they implement. Consequently, there is a need to express the commonalities and variability between products of a product family in terms of features. Unfortunately, technology support for the early aspect of a feature is currently limited to the requirements level. There is a need to extend this support to the design and implementation level as well. Existing separation of concerns technologies, such as AOP and SOP, may be of use here. However, features are not first class citizens in these paradigms. To address this and to explore the problems and issues with respect to features and feature composition, the authors have formalised the notion of features in a feature model. The feature model relates features to a component role model. Using our model and a composition algorithm, a number of base components and a number of features may be selected from a software product family and a product derived. As a proof of concept, the authors have experimented extensively with a prototype Java implementation of their approach.