Irina wrote:
I also have NOT used Bayes. Don't know how safe it is. Would I just submit a spam message and I don't have to anything else, or ham the same way? Not sure.
Some people have problems with Bayes, but many find that it does help a lot. It does require you to train it with both spam and ham, and if you enable autolearn, it may be worth setting the config item bayes_auto_learn_threshold_nonspam to 0 to avoid poisoning it with new spam that doesn't trip any rules.
In short, Bayes works by finding the trends in both spam and ham, then comparing each new message to those trends. It needs to be able to compare junk mail to legit mail in order to determine that, for example, "pills" is more likely to show up in spam, "the" is neutral, and "ninja" is more likely to show up in personal correspondence.
-- Kelson Vibber SpeedGate Communications <www.speed.net>
