Composing Models at Two Modeling Levels to Capture Heterogeneous Concerns in Requirements

Abstract

Requirements specification is initially scattered in numerous partial models (viewpoints), describing heterogeneous concerns (typically functional and non-functional ones). To define these concerns, requirements analysts prefer describing them separately with metamodels so that they can be properly identified, reused and tooled. The production of one unified view of requirements from separate viewpoints is a complex issue which requires a composition process working at two levels of modeling. At the meta-level, separate “of-the-shelf” metamodels allow defining either concerns or variation in the operational semantics. These metamodels have to be composed into a core metamodel, which captures the information and semantics needed for expressing and analyzing the requirements of a dedicated application domain (e.g. real-time critical systems, telecom services). At the instance-level, viewpoints are composed to produce a global requirements model, which has to be conformant with the core metamodel. Although the same composition mechanism is used for both levels, we emphasize in this paper the strong coupling between the two steps and the difficulty to make both compositions consistent with each other. We thus propose a process for dealing with two-level of composition. The process is illustrated in the context of a platform specialized for requirements analysis purposes.