Methods and Techniques for the Design and Implementation of Domain-Specific Languages

A risk of domain-specific languages is that their constructs are too high-level, preventing good coverage of the domain. While such DSLs typically support very concise solutions to a set of problems, they are rendered useless as soon as variations of the problem have to be solved that the language does not support. A manifestation of this problem can be found in the context of workflow description languages. Workflow is concerned with the coordination of activities performed by participants involving artifacts. For instance, a user registration workflow involves filling in a registration form, e-mail verification and administrator approval; the process of academic paper reviewing involves bidding, review submission, discussion and notification of authors. There are many DSLs that support high-level definition of workflows, including UML activity diagrams [Dumas and ter Hofstede, 2001], BPEL [Curbera et al., 2003], and YAWL [van der Aalst and ter Hofstede, 2005]. While workflow languages raise the level of abstraction from manually encoding to declarative descriptions of workflows, they have no mechanism to deal with coverage issues, i.e. what if a workflow follows a pattern that is not supported by the language? This raises the issue of maintaining coverage while introducing abstractions in a DSL: