Elmar Jürgens and Markus Pizka.
Variability models must not be invariant!
In Klaus Pohl, Patrick Heymans, Kyo-Chul Kang, and Andreas Metzger, editors, First International Workshop on Variability Modelling of Software-intensive Systems (VaMoS), number 2007-01 in Lero Technical Report, pages 171-176, Limerick, Ireland, January 2007. lero - The Irish Software Engineering Research Centre.
Abstract
Variability modeling techniques are used to specify variable aspects of members of a family of related software artifacts. Instances of variability models are then used to efficiently produce members of such a family. By making variability explicit, variability models determine implicitly the common properties among family members as well.
This partitioning of information into variable and invariant parts predetermines the reuse benefit obtainable from a variability model. In most current approaches to variability modeling, the decision between variable and invariant information has to be done in an up-front manner and is very difficult to change later on. However, the distinction between variable and common parts of members of a system family varies over time. Variability modeling techniques must thus be able to cope with changes to the variability models.
Since many variability modeling techniques do not currently support this, research is required to allow an evolution of the distinction between what is variable and what is invariant over time. This paper elaborates on the necessity, related work and possible approaches to tackle this challenge.
Keywords: modeling, software evolution, formal methods, software reuse
