Cool! Free wireless has arrived in Norwich (of all places).

It ssllooww! ~250 kb/s. But better than nothing. [Edit: even though it cuts out every so often; their web page says its an "outdoor network". Great, so we're all supposed to sit outside this winter in the cold and rain, are we!?] Mind you, my computer might be slow anyway because I'm using the composite extension to X with KDE's translucency effects, and this on a silly little laptop with a Micky Mouse graphics card. Ah, well. It looks cool at least.

I haven't posted here for ages. Partly because I'm just not interesting enough, and partly because I'm lazy.

I've done one or two interesting things since I last posted: put an initial version of the CURSUS search engine online: http://www.cursus.uea.ac.uk/

This project had been without a search engine for ages and now I've finally got round to working on one. It needs some refinement (problems with passing the terms between search result pages, displaying the relevant text in the results pages, and choosing a more realistic granularity for the Vulgate text). But, hey, its a start.

This also meant learning the excellent Swish-e (http://swish-e.org) full-text indexer which was fun (mainly because its easy :-D) and also integrating it into my XML Web publishing framework, Pycoon. (Yes, its yet another cheap rip-off of Apache Cocoon written in Python). This meant handling the long-dreaded upgrade from lxml (http://codespeak.net/lxml/) 0.9 to 1.0 and its altered API for unicode support. Bit of a pita but it worked in the end.

Should I have a go at publishing it as a SourceForge project? Maybe. I'd need to tidy the code up, write some basic docs and test it on platforms other than Linux/Apache2-mpm-prefork... I should also some production-type features like proper caching (at the moment it just caches the response stream and only updates it after the cache is full; it should cache the source streams as well and use modified dates and user options for better customisation), authentication (had some ideas for this) and some conditional logic processing once inside a pipeline so that responses can be customised for things like client agent and so that query string and post values handling can be improved.

Hmm..Actually, I might just do that.