Brad Templeton's article on the history spam has appeared on Slashdot and a bunch of other places. His article on what to do about it hasn't been referenced a much as the article on the history. The network of throttling servers that he proposes has a lot of nice properties. It also has the property of not depending on pattern recognition of spam content. I'm running Spamassassin on my mail server and Mozilla 1.3 with bayesian filtering turned on on the client, and I'm still seeing spam -- partially because I don't feel confident enough in Mozilla's bayesian filter to have it move spam messages aside.