I just received a reply from someone that consisted only of the following sentence: “Thanks for telling me”.
It got a BAYES_99 score.
Your system has likely auto-learned a few of the latest virus laden e-mails, which are usually in the form of a single sentence and a file attachment. The virus e-mails have many spam-like characteristics, which pushes them to the auto-learn spam score. Normal short e-mails don't have such characteristics, but they get the Bayes score, which is enough to push them over 5.
My system is setup to reject virus e-mails before calling SpamAssassin, thus avoiding this problem altogether.
Train the message as ham, and see how bayes handles the next one.
Steven -- Steven Dickenson <[EMAIL PROTECTED]> http://www.mrchuckles.net
