Section: New Results

Global Model Management

Participants : Hugo Brunelière [ contact ] , Guillaume Doux, Jean Bézivin, Frédéric Jouault.

As noted above, MDE generates a lot of artifacts that need to be handled if we want to harness the accidental complexity and if we wish MDE to scale up in the context of collective usage by teams of engineers of several models in different versions. The problem we tackled here is part of the "tower of models" grand challenge proposed by Robin Milner.

Problem 1. In our recent work we have found that what we must have more than single-shot transformations. We also need long chains of transformations or even complex graphs of transformations. For example in the collaboration with ILOG more than 25 transformations are being built and the engineer needs to control the corresponding complexity. With the notion of a megamodel we propose solutions to this problem. A megamodel is a model such that some of its elements are themselves models. From there, one problem is to make sure that transformations are composed correctly. To address this issue, AtlanMod worked on a type system for megamodels [27] .

Problem 2. As many companies realize that they may not rely anymore on one unique modeling language like UML, they face the problem of allowing their engineers to navigate between numbers of different models based on heterogeneous DSLs. This is an open problem that we are solving on a conceptual and practical basis. Until now all solutions to this problem have been specialized ad-hoc implementations [17] , [20] .