Paul Graham has a new piece on fighting spam titled, "Filters that Fight Back." Paul explains how spammers try to fool bayesian filters. He also suggests creating spam prevention software that follows every URL listed in spam to flood the spammer's http servers. I disagree on the http flood attack. It's to easy for abuse.

Kristian Eide compares two bayesian filtering programs, Popfile and SpamBayes. I'm a big fan of Popfile. I'm currently running around 98% spam id rate.