begin  quote
On Wed, 16 Jul 2003 17:51:46 +0200
Fredrik Jagenheim <[EMAIL PROTECTED]> wrote:

> 
> [1] Which I find utterly useless in spamassassin. The only thing the
> bayes filter in spamassassin does is to nudge things it already knows
> is spam a little further up in the Score list. I have yet to find a
> spam that sneaked into my inbox which has had a high Bayes score. And
> considering the low additional score even letters with Bayes_90 gets
> (per default) I find it hard that it would end up in the $SPAM bin on
> the Bayes score alone. Hum, I should write a script that mined my
> $SPAM bin for SpamAssassin scores and do some statistics on it.
> Although, I bet it's already written... :)

Well, I currently use a combination of both bogofilter and spamassassin
in a collective setup with spamc/spamd + bogofilter  along with
fetchmail+procmail..

So far they are piped through this recipie:

--
:0fw: bogofilter.lock
| bogofilter -u -e -p

:0e
{ EXITCODE=75 HOST }

:0fw: spamassassin.lock
* < 256000
| spamc

:0:
* ^X-Bogosity: Yes, tests=bogofilter
SPAM/bogo/.

:0:
* ^X-Spam-Status: Yes
SPAM/assassin/.
---
after this there is some work necessary to keep it up, I maintain a
"clean-spam" and an "uncaught" set of mailboxes as well, for future
reference and learning..

So far, before re-learning I usually do:
for i in assassin/* 
 do spamassassin -d <$i  | grep -v X-Bogosity  -  > TEMP/$i;
 bogofilter -s <TEMP/$i
done
sa-learn --spam --dir TEMP

for i in bogo/* 
 do spamassassin -d <$i  | grep -v X-Bogosity  -  > CLEAN/$i
done
sa-learn --spam --dir CLEAN



Perhaps a bit cumbersome, but it makes sure that the various filters
dont learn to sort spam based on the previous filters rulings ( which
would be rather counterproductive ;)

keeping a unified CLEAN dir is another set of work (mails in mh style
are reordered when emptied, so I usually just check the highest current
number then do a :
for i in TEMP/*; do mv $i CLEAN/$[${HIGHEST-NUMBER}+$[i}];done

Well, Some nice offtopicness into the delving world of spam, if somone
has a more efficient way to handle this, please let me know.

//Spider

-- 
begin  .signature
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to