On Monday 06 November 2006 13:59, itdelany took the opportunity to say: > I successfully run spamassassin with bayes filter on my site and is working > really really fine. But, this morning i noticed that an email i receive > everyday from my server (It's from LogWatch application, it sends detailed > log information from past events, like users logons and postfix statistics) > So i run > > # sa-learn --progress --ham a0f3773f-bc37-4eae-b912-5339ea06735d.eml > 100% [===========================================] 20.02 msgs/sec 00m00s > DONE > Learned tokens from 1 message(s) (1 message(s) examined) > > BUT the email keeps being caught as Spam by SpamAssassin, do i forgot > something?
SpamAssassin will not see all those messages as the same. However, one would
expect it to learn to recognise it as ham with time. Maybe it has, but the
negative score of BAYES_00 is too small to bring the total back under 5.0.
There should be an X-Spam-Status line enumerating the rules that hit.
> What exactly --forget do ? How can i assure this email will not be
> considered spam in the future?
--forget forgets about a previously learnt piece of ham or spam, like you
hadn't run sa-learn --ham at all.
A better approach however would be to skip running those messages through SA
at all, or to whitelist the sender address (read about whitelist_from_rcvd in
Mail::SpamAssassin::Conf(3pm) manual page. Also, is the server in
trusted_networks?
--
Magnus Holmgren [EMAIL PROTECTED]
(No Cc of list mail needed, thanks)
pgpDwXnXIMowd.pgp
Description: PGP signature
