[milters] Archive

Lists Index Date Thread Search

Article: 925
From: Taylor, Grant
Date: 2006-05-11 12:14:15 -0400
Subject: Re: Milter-link & SpamAssassin...

Removal...........: milters-request@milter.info?subject=remove
More information..: http://www.milter.info/#Support
--------------------------------------------------------

> This is where I'm confused, why not just have milter-link reject the 
> mail with -p reject. Or if you're still paranoid about loosing a 
> message, use -p quaratine instead.

I prefer to reject based on logistical tests, valid sender, valid MIME encoding, valid
date, etc.  I tend to tag on subjective things like a link that is in a black list that I
have no control over.

> Otherwise if you're using a milter/SpamAssassin combo already, why use 
> milter-link, if you're not going to reject or quaratine? SpamAssassin 
> can already do the URL checks (though more slowly I think) and tag it 
> itself.

The reading that I did indicated that SA only checked the URLs that were directly in the
message, where as Milter-Link would actually issue a connection to the URLs and follow
them back to the original server if they were redirects, something that (from my
understanding) SA would not do.  Thus it would be more likely that Milter-Link would find
a URL that was black listed somewhere than it would be fore SA.  As such I needed a way
for Milter-Link to do the tests and then for SA to act based on the results, thus the
header test that I submitted.

> If you do choose to use milter-link and SpamAssassin together, then 
> disable the URI BL checks in SpamAssassin, as they will be redundant and 
> slow you down.

I'm not so worried about the slow down.  I'm after /more/ ways to identify likely spam...
Rather I'm after more SA tests to aid in the determination of spamicity.  Thus hopefully
yielding a wider gap in SA scores for Spam verses Ham making it easier to more accurately
set the SA spam level for what is spam and what is not spam.  I would rather have the same
URL be caught by two or more (independent) SA tests (which operate differently than each
other) that will each raise the SA score.  I'm trying to go after the more tests that a
message fails the more likely that it is spam. Hence why I'm trying to put more tests in
place.

At present I'm running 7 different milters, having SA process about 4 different black
lists, with each test doing slightly different things in slightly different methods.  As
such I'm having most of the spam be caught and very little ham being caught.  My only real
problem is sending sights that do not accept the null reverse path "<>",
for which I'm white listing as need be.



Grant. . . .

Lists Index Date Thread Search