Feature and Feature Interaction Modeling with Feature-Solution Graphs

The architecture of a software system captures early design decisions. These early design decisions reflect major quality concerns, including functionality. We would obviously like to design our systems such that they fulfill the quality requirements set for them. Unfortunately, we in general do not succeed in doing so in a straightforward way. This is especially true for product lines that must evolve and/or must support variations with slightly different features. What we need is an approach that on the one hand can assess the impact of feature interactions and on the other hand can be used to generate different versions of a system dependent on the required features in such a way that all quality requirements are satisfied. This position paper1 is concerned with techniques to support this approach. In particular, we propose to use a rich featuresolution graph to capture the evolving knowledge about quality requirements and solution fragments. This graph is next used to pinpoint feature interactions and to guide an iterative architecture development and evaluation process. The structure of this feature-solution graph resembles that of the goal-hierarchy in goal-oriented requirements engineering [2, 3]. The solution fragments included in this graph have much in common with Attribute-Based Architectural Styles (ABASs) [1]. In principle, any kind of solution description will do. The approach to generating and evaluating architectures from a feature-solution graph is depicted in Figure 1.