April 15, 2003
comp.lang.python made flesh

Andrew Koenig wasn't able to make it to the UK Python Conference this year - his company has a travel ban due to the threat of terrorism. Anyone would think that America was unpopular, or something.

Anyway, Andy Robinson put together a short-notice replacement session - a panel discussion entitled "P2EE - is Python ready for the enterprise?" On the panel were Andy, Guido, Jacob Hallén, Laura Creighton and Marc-Andre Lemburg.

This discussion was comp.lang.python made flesh - good natured, rambling, frequently going wildly off topic, the occasional total non sequiter, that sort of thing.

One of the threads of discussion was about Python advocacy. Andy and the other Python-in-a-tie people are really big on this. To them, Python is strategic, it's a 'platform'. They want it to have 'market share', they want CEOs to have heard of it, and they want them to have heard good things about it. Me, I can't get excited about this. I sympathise with their position, but Python, to me, is tactical - it's a tool. It's the best goddamn tool I've ever had, and it's a beautiful and elegant tool, but it's just a tool nevertheless. I don't care if CEOs have heard of it. So long as the user base stays large enough to keep the 'platform' viable, that's all I really care about. Which is probably lucky - Python in unlikely to ever have a marketing budget to speak of, so Python will probably only ever grow organically.

There was then a quick discussion of the difficulties of application deployment, and distutils' unhelpfulness in this regard. I didn't really follow this with my full attention - as a way of installing libraries, either pure Python or with C extensions, distutils just works for me. As for application deployment, well, that is just plain hard in general. My needs are very simple, and I've been happy using py2exe and Inno, though I'm thinking of switching to NSIS. But if you want cross-platform scriptable installers, then I'm not sure what exists. Distutils, though, was designed to simplify module and package distribution rather than application deployment, or so it seems to me. Marc-Andre suggested that extending it to support application deployment is trivial enough that everyone could use it, but I think that he was forgetting that most of us don't have a brain as large as his.

We then talked a little about web application frameworks, and the complete and utter lack of a de facto standard in this area. This has been pretty much reiterated on c.l.py now. I myself would be interested in using a Webware/Cheetah combination, but I'm rather stuck in Java-land, where J2EE/JSPs/Taglibs rule the roost. I'm interested in Struts - pure JSPs are an invitation to the Magic JSP anti-pattern discussed in Bitter Java. But I'm also very interested in Anthony Eden's JPublish project. It uses Velocity for its templating, which is *much* nicer than JSPs, and for the action scripting and business logic, you can use Java or any of the BSF languages, which gives you access to Jython. I'll blog more about this later, because this is a lovely approach.

Another issue discussed - documentation. Andy Robinson felt that Python should provide better tools for the production of documentation. Python's docstring approach is a good one for 'embedded' documentation, and for the generation of API reference documentation. reStructuredText support will just be the icing on the cake. As for the rest of the documentation, well, as I pointed out, this is not a technological problem, but a sociological one - people have always failed to write documentation, and they probably always will. Good tools exist, but people would rather do other things.

On the off chance that anyone is interested, Duncan Grisby has put up some photos taken at the conference, many of the ones from Wednesday taken at this session. It's interesting to put faces to names...

Posted to Python by Simon Brunning at April 15, 2003 04:05 PM
Post a comment

Email Address:



Remember info?