Exploiting the ASM method within the Model-driven Engineering paradigm

Model-driven Engineering (MDE) [3] is an emerging approach for software development and analysis where models play the fundamental role of first-class artifacts. Metamodelling is a key concept of the MDE paradigm and it is intended as a modular and layered way to endow a language or a formalism with an abstract notation, so separating the abstract syntax and semantics of the language constructs from their different concrete notations. Although the foundation constituents of the MDE approach are still evolving, some implementations of the MDE principles can be found in environments like the OMG MDA (Model Driven Architecture) [6], Model-integrated Computing (MIC) [7], Software Factories and their Microsoft DSL Tools [1], etc. We believe that the MDE paradigm can gain rigour and preciseness, necessary expecially for model analysys purposes, from the integration with formal approaches. We, in particular, try to address here the problem of defining the operational semantics of metamodel-based languages in the MDE approach. Currently, metamodelling frameworks lack of a way to specify the semantics of languages. The OMG metamodelling framework, for example, provides two main standard techniques to define languages: metamodels and UML profiles. Building metamodels is a first-class extension mechanism which requires the definition of a new metamodel or an extension of an existing metamodel based on the MOF (Meta Object Facility) metalanguage i.e static class diagrams and well-formedness rules written in the OCL (Object Constraint Language)[8]. The UML profile mechanism is a non-first-class extension mechanism, i.e. it does not allow for modifying the existing UML metamodel, but it requires the specification of UML extension elements (stereotypes and tagged values) and the definition of new constraints (well-formedness OCL rules) specific to a target domain. For both extension techniques, the way to define the abstract syntax (i.e. the MOF-compliant metamodel) and static semantics (the OCL constraints) of a new language are well established. However, the OMG framework does not have any standard and rigorous support to provide the dynamic (operational) semantics, which is usually given in natural language.