On 18 mars 09, at 09:28, Bergel, Alexandre wrote: Actually having FAMIXFunction in the Core is not a problem, as long as it only contains metadescribed attributes and classes. As Doru says, the goal is to have a generative and auto-testing core through Fame export/generation. Then, for example, methods belonging to FAMIXFunction but not metadescribed should be class extensions defined in their own package. However, I would like to keep the FAMIX-Core minimal as I believe it is quite complete. Here is the current structure of Famix3 packages Famix-Core: meta-described entities and attributes Famix-Implementation: mostly generic class extensions, like helper methods, which can not be meta-described Famix-Extensions: quite generic stuff, but does not belong to core, like metrics, the Visitor... Famix-Smalltalk: class extensions and classes specific to Smalltalk model (for example, FAMIXMethod is extended with a #categories attribute) Famix-Java: class extensions and classes specific to Java model Famix-File: an extension to describe a file structure etc. My recommendation: do your stuff in your own package, dont think about the metadescriptions, you can integrate them later once your model has settled. It will be easier. -- Simon

Hi, I did not want my previous mail to sound like scolding, I just wanted to start the discussion on what the expectations are :). Simon summarized it quite nicely. However, coming back to FAMIXFunction, I think that it should be in the core. The reason is that we sell FAMIX-Core as the core of a language independent meta-model. So, by default the basic concepts should be in, and FAMIXFunction is quite important for languages like Lisp or C/C++. Cheers, Doru On 18 Mar 2009, at 10:05, Alexandre Bergel wrote: -- www.tudorgirba.com "Some battles are better lost than fought."