Hi Tom, I have not been using Bayes but have been considering how to properly implement it on my gateway. I had some ideas about using output from SA as input for Postfix smtpd/header/body rejections. At the end of July Jason Jordan put together a Perl script that would tail the "maillog for "554 Service unavailable" messages as generated by Postfix RBL's. After $definable of these are generated by a source, they are firewalled."
I was thinking of a similar idea. However, I would want to probably acquire the information from my Procmail log and use it to modify the smtpd_???_restrictions and/or header/body checks. The modification would need to take place after "$definable" occurances. I still doubt how effective this would be and I am leary as the MTA rejects are *final*. I get tempted as rejecting at the MTA just saves so much resource for real stuff. I get really concerned about FPs with an smtpd reject. That thought concerns me more than Bayes. --Larry > -----Original Message----- > From: Tom Meunier > Sent: Sunday, October 12, 2003 10:03 PM > So I'm loving the rules http://spamhammers.nxtek.net/ that > Jennifer Wheeler wrote , but I'm up against a few > (philosophical?) questions, and would like to invite discussion. > > I've noticed that about 95% of the time when these rules are > hit, they're listed as BAYES_99. In this case, should I even > bother? I guess they ensure that the 5% that are not > BAYES_99 have more of a chance of getting fed back into Bayes > because of it, and that's A Good Thing. > > I've also considered dumping spammers at the MTA level before > they are even passed to SpamAssassin, using a homebrew RBL or > something. The good part of this is that it would save > traffic on my Postfix/SA gateway box, then my AV box, then my > mailserver (drive space here, too) itself. The bad part is > that these spams would then never get learned by Bayes. I do > blacklist defunct users at the Postfix level, and that > actually cut my spam traffic by over 50% by itself, but again > this stuff never gets learned in Bayes. > > Has anyone else juggled these questions? > > -tom ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. SourceForge.net hosts over 70,000 Open Source Projects. See the people who have HELPED US provide better services: Click here: http://sourceforge.net/supporters.php _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
