Well, I started giving a run at DjangoPaste. I was hoping it would be more straight-forward than CherryPy, but maybe it's actually a bit worse. Lots of things are global -- configuration is loaded and used on import, not at runtime. This is the hardest part of Pastifying an application/framework. Right now I only have an entry point for projects, when really I want to access individual applications, as I think Paste makes projects unnecessary. The only fundamentally complicated part of that is the admin screens, which are applications that know about other applications. I think that the admin application will have to be a middleware of sorts (composite app, in particular).

But, even if Django might actually be a bit harder, the primary ticket that I opened is already in the process of being fixed the morning after, so there's good momentum. And really I'm still very fuzzy about how well CherryPaste will work, because CherryPy isn't very expansive, and even if I can get the framework working that doesn't mean each application will work. So ultimately I'll really just be targetting TurboGears, as that's a set of applications that work similarly. Django involves things like database configuration up front, so if it works for one app then it is likely to work for all of them.

Really this whole process of adapting projects to Paste started with a patch to Trac. That went really easily (even though there's more I'd like to do with it), and it made me feel optimistic about this assimilation process. And that was based on the ability to monkeypatch and otherwise force compatibility separately from the original package. I always figured that would be transitional, that it would be good to bring that support into the original package directly, but I like the idea of not having to negotiate each conversion. Now I'm a little less optimistic about that. But we'll see how it goes.