From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Masover Subject: Re: why does reiserfs list get so much spam? Date: Fri, 16 Sep 2005 13:51:39 -0500 Message-ID: <432B143B.8030808@slaphack.com> References: <432A9BFD.10700@mch.one.pl> <20050916104129.GA11025@kruemel> <432AA409.5070006@interia.pl> <432AAA38.9000102@mch.one.pl> <432B0DD0.80206@slaphack.com> <432B0FAC.4050905@mch.one.pl> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <432B0FAC.4050905@mch.one.pl> List-Id: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Tomasz Chmielewski Cc: =?ISO-8859-2?Q?=A3ukasz_Mierzwa?= , reiserfs-list@namesys.com Tomasz Chmielewski wrote: > David Masover schrieb: > > (...) > >>> If You will look in the headers of messages that You get from this list >>> You will see that there is spamassassin running on thebsh.namesys.com, >>> it's just that it is not configured good enough. >> >> >> >> Can spamassassin be configured "good" enough? >> >> I use dspam: >> >> http://www.nuclearelephant.com/projects/dspam/ >> >> There are some articles about why dspam has a fundamentally better >> design than spamassassin, and why in general statistical filters beat >> manual-rule-based ones like spamassassin. > > > Yeah spamassassin can work extremely well. > > I guess these articles are based on the quality of spamassassin which > checks spam from this list? :) > > And it's not really true that spamassassin is a manual-rule-based filter > only. Right, but the statistical/learning component of spamassassin is just that -- a component, to be combined with razor/pyzor, manual rules, and anything else they can think of. I think dspam does a much better job at being a statistical filter, and that's all it does -- and that's all it needs to. Some people have reported 99.997% accuracy from dspam, beating humans. Anyway, the articles are about the principle of the thing. A statistical filter will beat a manual one every time, because it's faster and better at coming up with rules, and you don't need to update your definitions to start filtering the new spam -- just train on two or three mails, and you're done.