Re: [SLUG] Training bogofilter problem

From: Jason Boxman (jasonb@edseek.com)
Date: Sat Sep 09 2006 - 18:06:07 EDT


On Saturday 09 September 2006 17:57, Paul M Foster wrote:
<snip>
> I have to echo this. I was a fan of bogofilter when it first came out.
> But its databases will grow into the megs if left unchecked. Moreover,
> you *must* train bogofilter with a half-and-half mix of spam and ham.
> But for me, list traffic and traffic from family bypasses all my other
> email filtering. So I don't want it going to bogofilter and getting
> graded; there's no point in chewing up CPU cycles that way. But without
> that, 90% of my other mail is spam. You can't train bogofilter on just
> spam, or it will think everything is spam. SpamAssassin is by far a
> better solution. On top of that, learn procmail.

SpamAssassin is, of course, more of a spam scoring engine that pulls from many
sources, making it more resilient to different attacks. The cost is in CPU
time and memory, not to mention keeping up with all those different spam
solutions you plug into SpamAssassin.

These days, I've found a mix of greylisting and rejection based on invalid
HELO strings and bad rDNS ranges (i.e. 12-55-234-89.foo.dyn.blah.com) coupled
with dspam to work reasonably well. greylisting has lost its effectiveness
over the past year, though.

-- 

Jason Boxman http://edseek.com/ - Linux and FOSS stuff

----------------------------------------------------------------------- This list is provided as an unmoderated internet service by Networked Knowledge Systems (NKS). Views and opinions expressed in messages posted are those of the author and do not necessarily reflect the official policy or position of NKS or any of its employees.



This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 17:02:28 EDT