Multi-Version Decision Propagation for Configuring Feature Models in Space and Time
Tobias Heß, Simon Karrer, Lukas Ostheimer · 2024
Real-world feature models are typically too large and complex to be configured manually. In practice, configuration tasks are, therefore, accomplished by employing interactive configurators. After each explicit feature selection or deselection by the user, these configurators use decision propagation to detect features that are implied by the current partial configuration and, consequently, select or deselect them accordingly. This way, the configuration remains valid throughout the configuration process. However, valid configurations may become invalid when then underlying model changes due to model evolution. As one is potentially interested in retaining a configuration for multiple versions, for instance, in testing or certification applications, this prompts the question on how to configure for multiple versions at once. In this work, we introduce multi-version decision propagation which allows to interactively configure on multiple model versions at once. Our prototype adapts the set of possible versions to the current configuration but also allows users to configure for a fixed set of versions.