>Afier,
>
>I think this depends of how sure you are of your framework. Me, being cautious, I use a copy of my framework in each new application.
>
>It gives me the ability to extend the functionality of the framework without having to think of previous applications.
>
>On the other hand, I might very be creating a real version maintenance headache for myself :).

In the structural world, I had a framework in FPD. My policy was to keep temporary experimental versions of framework things in app directories, and if a novelty survived testing and a couple of months of life, it was copied over the original in the (common) framework directory.

In OOP world, I'd have empty subclasses of most of the classes of the framework in a classlib in the app directory, and using them as intermediary, change what has to be changed in this specific case, and, once it lives, see which changes are candidates to be propagated back into original classes (i.e. if other apps would benefit of the new behavior). In the current phase, all apps are using the framework (or what is ready of it) off the same directory

Someone was looking for a way to copy classes into empty classlibs just yesterday.