TurboGears - off the Rails?

Well, it looks like Python is getting another framework with at least as much "media" impact as Rails - yes, Django was launched recently, but I don't think it had a 20-minute screencast, which is almost a perquisite for it being "über cool" these days.

Like all frameworks, TurboGears looks nice at a first glance. But I'm a skeptic where it regards frameworks (and there seem to be entirely too many of them right now), so what caught my eye was that it goes about its business in a very Pythonic way, re-using CherryPy, MochiKit and the Kid templating engine.

All of these are moderately popular in the Python community (and CherryPy seems to have a lot of traction in some quarters), so there's a lot of potential critical mass here.

Sadly, I don't go in for any of them separately, and one of the reasons I went with Snakelets was that I wanted a small, cohesive whole to deal with, preferably wrapped around an HTTP server I could optimize the hell out of - and one that does not require any sort of super-user access to install or run.

Which means I'm not crazy about TurboGears's installation methods either, since I'd much rather have a self-contained file tree that I can move from machine to machine without any requirements but Python, and Snakelets gives me precisely that. Plus, it runs fine in Python 2.3 (apparently TurboGears may have some issues, and I don't fancy upgrading my Macs' stock Python, nor installing a Fink or DarwinPorts version).

If you're a regular visitor, you may have noticed my (now hopelessly outdated) Python framework overview, which will give you some more background on my choice.

However, every time a new release of Snakelets comes out it tends to completely break all my code (I like clean URLs and try to masquerade all the templating behind URL prefixes, so I tended to overuse request objects in a way that is now deprecated), and it has only one developer (Irmen de Jong, who does a very nice - if under-appreciated - job).

And, of course, the fact that the screencast got a simple Wiki off the ground in 20 minutes (whereas I've been puttering about mine for, what - a year now? more?) was somewhat sobering.

Then again, I'm going for versioned filesystem storage (not a database) and I'm not that much of a hurry.

As always, it's nice to have options - and this looks like a promising one.

I'll just let it settle a bit first, OK?