jdow wrote:
From: "Rick van Vliet" <[EMAIL PROTECTED]>
Hello. New to the list, I have a question that I hope isn't "too newbie".
Running SA 3.1.2 with a qmail server for a small (50) group of users.
Vpopmail handling virtuals, and procmail.
(auto_whitelist is disabled)
I have one user who is getting creamed and no matter how much we do
sa-learn --spam...on the IMAP folder we move his spam into...this
user's mail somehow gets through with low scores, and he's averaging
60 spams a day, total, with FORTY of those that actually get to his
inbox.
Other than changing his email address, how would I teach SA that this
is spam?
Thanks,
Rick
Return-Path: <[EMAIL PROTECTED]>
X-Spam-Checker-Version: SpamAssassin 3.1.2 (2006-05-25) on
mail.vanmorel.com
Eliminate his Bayes filter and start over? On the other paw, there is
NO Bayes score on his mail as noted below. Are you learning the spam
under his account so that his Bayes is the one refreshed? Or do you
have a rule that is turning off Bayes for him altogether in his
user_prefs?
{^_^}
Hmm.
1) With vpopmail(virtual) under qmail, there's one user_prefs that
handles all virtual users.
*use_bayes 1*
*bayes_auto_learn 1*
2) That said, I better look at how I set up Bayes.
sa-learn /runs/ when I run it...just not sure why it isn't being looked at.
Another virtual users's headers look like this, and it does have an
"autolearn=":
X-Spam-Checker-Version: SpamAssassin 3.1.2 (2006-05-25) on
mail.vanmorel.com
X-Spam-Level:
X-Spam-Status: No, score=0.0 required=4.7
tests=HTML_MESSAGE,MIME_HTML_ONLY, MSGID_FROM_MTA_HEADER autolearn=ham
version=3.1.2
Delivered-To: [EMAIL PROTECTED]
(This user gets very few spams -- 98% get tagged properly.
But this one also has no Bayes score either, does it?
Looks like my Bayes component might be suspect?
I ran the install using the CPAN/perl method. Not sure now what kind of
configure was involved.
I'll look at the INSTALL, and see what I missed in the configuration.
(if I can find that ;)
Thanks,
rick
