Wow. So, just days after Paul Graham put up his Plan for Spam page on a possible Bayesian filter for spam, people have already started implementing it. This one is for Perl and qmail, but I’m sure I’ve seen another somewhere. And I’ll bet there’ll be many many more.
Bayesian analysis is one of those concepts that is so counter-intuitve I find impossible to understand it for more than five minutes. Must try harder.