The AmmA ToolBox
ATL, AMW, TCS, MoDisco, and AM3 are among the most important Eclipse.org components produced by the AtlanMod team. However there are also other components and a lot of functionalities, examples, and use cases made available and necessary to express solutions to many problems. The whole set of contributions composes the AmmA platform  .
One difficulty of developing research prototypes intended also to become practical industrial tools is the necessity to stay partially aligned with the standards of the market at the time the research is developed. We have managed to keep the ATL prototype sufficiently synchronized with the OMG QVT evolving recommendation without prejudice to the conceptual rigor of the proposal. In the domain of metamodeling we have similarly kept aligned with the OMG MOF and with the Eclipse Java-based Ecore by defining an independent solution called KM3 with good semantic properties described in  . The price to pay to remain reasonably independent from standards and industrial practical applications is bearable since today we have one of the best self-contained modeling frameworks and this is mainly due to the simplicity of the KM3 solution.
The success of the ATL language allowed building one of the first libraries of model transformations as part of the Eclipse.org/m2m/atl project. This library received contributions from many parties. It consists of a rich set of components, at various levels of maturity. One of the multiple conclusions drawn from this experience is that the number of modeling artifacts (terminal models, metamodels, transformations, etc.) involved in a given project is quite important. This creates an accidental complexity that may prevent MDE to scale up in practical industrial environments. In order to show that MDE is able to deal with complex situations beyond toy problems, new solutions need to be defined. The experimental construction of tools to deal with complexity need to be engaged as soon as possible. This is why we started the study of the megamodel management tool to handle complex situations when there is a high number of DSLs involved with a lot of semantic relations between these DSLs.
In retrospect, one of the main contributions of the second period of research in the ATLAS modeling group has been to show the power of the transformation paradigm over more classical direct execution schemes, based on Java or other general purpose languages for example. Through many examples we have demonstrated that many problems could be expressed as chains of information transformation. There is an obvious gain when we can express a problem as a chain of transformations in a declarative modeling language over a similar expression as a set of imperative statements in a general purpose language. However we do not consider ATL as the definitive model transformation language. On the contrary we view it only as a representative of a broad family of such languages yet to be precisely identified. Fortunately parts of the AmmA toolbox may be reused in the implementation of these new languages. Our team has also been the first one to show the deep practical impact of Higher-Order Transformations (HOTs) in practical modeling environments. This has been made possible by the fact that ATL is a true DSL, and that ATL programs may be considered as models conforming to a given metamodel  .