Can Conditional Preferences and *CP-net Concepts Enhance Feature Models?

Variability in software systems is a key concept in mastering complexity. Most software systems exhibit functionalities that can either be implemented in several different ways or may be options that can be chosen late in their design, depending on the deployment conditions. Expressing these forms of variability attracted a lot of attention since several years, leading to a predominant expression media: Feature models. Feature models can express the mandatory/optional status of a feature, as well as alternative realizations of a feature, exclusive of each others or not. Though extensions of feature models have been proposed to express more properties, they are still limited in their ability to capture complex configuration problems, especially those involving extra functional properties. Lines of research have proposed extended feature models with non-functional attributes and constraint-solving approach to reason about them. Yet, these approaches appear too restrictive and inflexible to cater for really complex SPL. In this paper, we rather propose to extend feature models with conditional preferences concepts from the multi-criteria decision-making field, and more specifically on relationships introduced in the graphical modeling approach of the * CP-net family of models.