In the context of MDE (Model-Driven Engineering), our objective is to define the semantics for a given DSL (Domain Specific Language) either to simulate its models or to check properties on them using model-checking techniques. In both cases, the purpose is to formalize the DSL semantics as it is known by the DSL designer but often in an informal way. After several experiments to define operational semantics on the one hand, and translational semantics on the other hand, we discuss both approaches and we specify in which cases these semantics seem to be judicious. As a second step, we introduce a pragmatic and instrumented approach to define a translational semantics and to validate it against a reference operational semantics expressed by the DSL designer. We apply this approach to the XSPEM process description language in order to verify process models.

One major advantage of executable models is that once constructed, they can be run, checked, validated and improved in short incremental and iterative cycles. In the field of Software Process Modeling, process models have not yet reached the level of precision that would allow their execution. Recently the OMG issued a new revision of its standard for Software Process Modeling, namely SPEM2.0. However, even if executability was defined as a mandatory requirement in the RFP (Request For Proposal), the adopted specification does not fulfill it. This paper presents a critical analysis on the newly defined standard and addresses its lacks in terms of executability. An approach is proposed in order to extend the standard with a set of concepts and behavioural semantics that would allow SPEM2.0 process models to be checked through a mapping to Petri nets and monitored through a transformation into BPEL.

No abstract