to expand on Camille's question, we are having fun here modeling cobol programs. So currently for example, cobol programs are modeled as modules which fits our need for now, but is kind of uncomfortable semantically: Modules are scoping entities they were not planned to have behavior. Of course Cobol programs do have behavior (it can be modeled as functions inside the module). Also if we model shellscripts as modules, we will have modules that can be cobol programs or shell scripts. Not very satisfying, again. And how to model cobol program calls from the shell scripts? The closest thing would be Invocations, but they are really tailored to method calls (e.g. with a sender and again they won't work with modules but only behavioural entities) So in summary, I think there are two questions: - the current Famix model targets essentially OO languages (more specifically Smalltalk and java), how can we best model procedural languages - how to strike the right balance between a generic metamodel and a detailed one ? nice questinos, we though you could be interested in discussing ... nicolas On 15/10/12 14:57, wam wrote: