David Goldsmith wrote: <sniped>
We are running SA 3.1.0. Reading this thread today, I just found the SARE_STOCKS ruleset. I updated the rules_du_jour script and pulled down the ruleset. Have received some messages already that are being caught.Some others are making it through with scores of 6.7 - 6.9. Our threshold is 7.0. I have two issues/concerns with that: 1) When I run "sa-learn --spam <file>", I'm getting the following warning: $ sa-learn --spam 1149292740.12607.iceman12.giac.net Possible unintended interpolation of @1 in string at (eval 4924) line 1. Learned tokens from 1 message(s) (1 message(s) examined) 2) After running sa-learn, running the messages through spamc again doesn't show any improvement in the score. Is this normal? Thanks, Dave
How many messages have you trained? You'll need 200 each to get it going, and I recommend at least a thousand of each to really get it going. That said, I don't run autolearn. I have the privilege that we receive the same kinds of mail pretty much across our user base. Only the accounting department gets 'spam' looking messages. Most of our suppliers implement SPF, so whitelist_from_spf is a blessing (3.1.x and above.)
-- Thanks, JamesDR
smime.p7s
Description: S/MIME Cryptographic Signature
