Computer scientists are increasingly aware that modern general-purpose programming languages are too low level--not with respect to the underlying hardware, from which they are increasingly distant, but with respect to the problem domains where they are applied. We are thus witnessing the growing pains of a whole new breed of languages: modeling languages.

More specifically, in order to really benefit from further abstraction from hardware and general-purpose computing mechanisms, a number of domain-specific modeling languages are being created and studied. This has led to research on what can really be done with such an approach, what gains are possible, and what methodologies should be used. This paper is an excellent example of this trend: through the use of nontrivial case studies--in this particular case, building mobile device applications--we can gain a better understanding of the capabilities of such an approach.

After a motivating introduction, Mannadiar and Vangheluwe give a firm grounding to their approach via a transformational approach (particularly to statecharts). Along the way, we learn about the many benefits of modular synthesis--the modern terminology for writing a modular code generator. The gain is real: a lot of software artifacts that are traditionally done by hand can be generated automatically. At the end, the authors get more speculative and discuss how nonfunctional requirements, such as performance requirements, could be explicitly tackled in such a setting. This is also a clear improvement over current methods, as such nonfunctional requirements are usually relegated to documents that are only targeted at humans.

This paper is appropriate for anyone interested in a snapshot of the current state of research in domain-specific modeling and its use in application synthesis. The one disappointing aspect of the work is that the authors only use rather simple code generation methods. They do not seem to be aware of a decade of work in this area, leading them to make some disputable claims about “polluting the generator’s code.”