Well, it's been a few days since I gave my presentation (hopefully up with the other PyCon papers soon). Those are just the slides -- I'm planning on filling it out with more text and based on some of the reaction I've gotten. Slides aren't really that interesting... maybe when the audio is up.

Anyway, the presentation went pretty well. I think it was one of the best attended tracks of the conference. Maybe that's just because the PyWebOff presentation was really popular and people stuck around ;) -- or because it was early enough in the conference that people hadn't organized many competing open spaces. The response was good as well -- I think Phillip (who wrote the WSGI spec) even said it was the best presentation of the idea behind WSGI that he had heard; that's cool.

What I really wanted to sell was WSGI and WSGI middleware as a basis for different web framework developers to cooperate and share, and I'll try to make that argument again in the paper because I'm really interested in getting framework developers on board at this point (not so much framework users). Once we get our house in order we can get back to users (and none of this work makes anything worse for users).

We had several open spaces during the conference related to WSGI. The first turned into a bitch session about the problem with Python web programming -- but that's not very constructive. I had a chance to show a few people the code in WSGIKit, hoping that they'd get ideas about how they could refactor their own code in the same style. A couple people were particularly interested -- JJ Behrens (of Aquarium) was particularly interested, and we discussed a bunch of things -- unfortunately Aquarium might not be a good fit, since it uses coroutines which makes it a little funky. Jacob Smulyan (of SkunkWeb) expressed interest in the sprint, but I didn't see him at the conference... maybe he didn't make it. AMK (of Quixote) has been refactoring much of Quixote based on this (maybe in a branch), and was planning on experimenting with the result on his own site -- I think Quixote is one of the more decoupled frameworks anyway, so it's probably a bit more natural.

WSGIKit in turn has been making some good progress I think. It's downright useable (should I be proud of such a low bar? ;), with some new additions: configuration, exceptions, error reporting, auto-reloading, and a fairly easy-to-use startup script. Still needs more documentation -- I'm not sure if I should start with programmer documentation of the middleware (which is mostly in the docstrings anyway), or end-user documentation of the stack, or refining the to-do application as an example.

Once I feel confident of the error reporting I hope to start putting it in place at work -- to put something into production I'm much more worried about an error that I'm not aware of than I am of errors in general.