Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 22 Sep 2003 18:43:51 +0200
From:      Brad Knowles <brad.knowles@skynet.be>
To:        Marc Ramirez <marc.ramirez@bluecirclesoft.com>
Cc:        chat@freebsd.org
Subject:   Re: What are people using for MUA's nowadays?
Message-ID:  <a06001a14bb94d587ba6c@[10.0.1.2]>
In-Reply-To: <20030922121327.R335@www.bluecirclesoft.com>
References:  <20030922104213.L335@www.bluecirclesoft.com> <28213.216.195.235.103.1064243311.squirrel@webmail.gigguardian.com > <20030922121327.R335@www.bluecirclesoft.com>

next in thread | previous in thread | raw e-mail | index | archive | help
At 12:22 PM -0400 2003/09/22, Marc Ramirez wrote:

>  I'm probably most interested in setting up a Bayesian filter for now.

	Bayesian filters are one piece of the package, but should not be 
used alone.  For one thing, they need input of several hundred 
(preferably several thousand) recent spam/ham messages, so that they 
can be trained on what your particular mail traffic looks like.  If 
you can't give them that amount of input data, then they're not very 
useful.


	In my experience, you're better off starting with a rules-based 
scoring filtering system where things have been pre-assigned certain 
weights, such as SpamAssassin.  In here, you put all your black 
lists, and everything else you can (black lists just become another 
input to the score, helping to raise or lower the chance that the 
message will be recognized as spam).

	You can add to this a Bayesian-style learning/adaptive filtering 
system, and you should look closely at various options, including the 
Bayesian mode of SpamAssassin, or crm114 (which claims to get better 
performance than SpamAssassin on smaller sample sets), and I'm sure 
there are many others.

	To this picture, you should also add greylisting and a 
message-digest validation mechanism such as DCC, Razor, or Pyzor (I'd 
use at least DCC plus one other, and not just one of the three).

	Then you also need to incorporate anti-virus scanning.  I'm 
seeing more than two-thirds of my traffic right now being nothing but 
w32.swen.a@mm virus crap.

>  Anyways, thank you very much for your input. I'm gonna look at
>  SquirrelMail, too (for different reasons).

	For webmail solutions, I recommend TWIG.  It's PHP-based, but 
doesn't make nearly so much use of Javascript as Horde/IMP, and it 
seemed to work better for us than SquirrelMail at the large Belgian 
ISP that I used to work for.

-- 
Brad Knowles, <brad.knowles@skynet.be>

"They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety."
     -Benjamin Franklin, Historical Review of Pennsylvania.

GCS/IT d+(-) s:+(++)>: a C++(+++)$ UMBSHI++++$ P+>++ L+ !E-(---) W+++(--) N+
!w--- O- M++ V PS++(+++) PE- Y+(++) PGP>+++ t+(+++) 5++(+++) X++(+++) R+(+++)
tv+(+++) b+(++++) DI+(++++) D+(++) G+(++++) e++>++++ h--- r---(+++)* z(+++)



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?a06001a14bb94d587ba6c>