> Just posted on slashdot: a Bayesian approach to the problem that reports > to have rates of 0.5% on false positives and 0% false negative: for one recipient. try to make the same table serve more recipients, and the number of false positives quickly goes way up. Keith