#11 Work on QVTc2ATC to support QVTr2QVTc execution

[E. Victor Sanchez]

It has been said that the current search strategy seems fairly brute force; so it might be simple and might work and provide a reference for a more optimsed approach.

It seems reasonable to think about a useful set of ATC procedures supporting alternate search strategies that can be exploited either by manual override, meta-model-driven heuristics, model-driven profiling. Well worth trying to formulate an API and separating development between language translation and language implementation. In due course the API could move across the boundary to be ATC 'byte' code.

This implies a fair amount of QVTc2ATC development.

However let's do things in the right order; an accurate QVTr to QVTc to ATC to user behaviour is most important. With that in place we can write optimisations in QVTr or something above and use the slow accurate tool chain to produce a steadily faster tool clain.