Separation of Concerns in Translational Semantics for DSLs in Model Engineering

Development of Domain Specific Languages (DSLs) in the context of Model Driven Engineering is gaining more and more popularity. As evolution lies in the heart of every software system, the major requirement for DSLs is that they should be modular and resilient to changes. MDE-based DSL frameworks should enable a modular specification of language translational semantics and the composition of the modules into languages. Ultimately, the availability of such techniques should make the DSL development faster. Separation of concerns is a sound software engineering principle used to obtain better modularity, reusability, and adaptability of systems. However, this principle must be supported by proper tools that allow the separation achieved at a conceptual level to be preserved in the language specification. In MDE, the mainstream tools for specifying translations are model transformation languages. In this paper we evaluate a class of model transformation languages regarding their applicability for capturing the translational semantics of DSLs in a modular way. We found that the concepts in the domain of translational semantics significantly mismatch with the language constructs of the transformation language. We suggest that this problem may be better approached by a domain-specific transformation language.