Für Studenten

Für Unternehmen






RWTH Aachen

On the Pragmatics and a Constraint-based Semantics of Multilevel Modeling

In this talk we discuss practical and formal aspects of the modeling of sets of sets of objects. The modeling of sets of sets is an important issue in domain modeling, because it arises naturally in expert domains. The modeling of sets of sets has been discussed as multilevel modeling in the past. Concrete modeling constructs like powertypes and clabjects have been introduced for it. Unfortunately, concepts from meta modeling frameworks have been accidently identified with concepts of the modeling process in the past. This has led to confusion in important issues of the semantics and pragmatics of multilevel modeling. This status of confusion is still not overcome, which can be seen, e.g., by a closer look to the current UML specification. This talk aims at providing keys to the precise understanding of multilevel modeling constructs and their exploitation in tools and technologies. We provide an explicit and precise formulation of the necessary constraints for set construction as a so far missing link in the discussion of multilevel modeling. We use the well-established constraint language OCL for this. We carve out the crucial aspect of the intuitively intended meaning of powertype and clabject constructions, i.e., in how they impose M2/M1/M0-overspanning constraints that are robust against certain M1-level model updates. We introduce symbolic viewpoints on model manipulation that explicitly overcome the M1/M0-level divide of the current mainstream viewpoint. These symbolic viewpoints also achieve a systematic generalization of the schema/object evolution/migration scenario that is practically so important to type/instance-evolution at all internal levels of the combined M1/M0 level. This way we overcome certain inconsistencies and semantic fallacies in the current discussion and want explain the promise and the advantages of constructs like powertypes and clabjects for emerging tools and technologies.