[svlug] Spamassasin and Spam

[Karsten M. Self]

on Fri, Dec 19, 2003 at 09:21:29AM -0800, Drew Bertola (drew at drewb.com) wrote:
> On Thu, 2003-12-18 at 14:44, Karsten M. Self wrote:
> > on Thu, Dec 18, 2003 at 03:16:56PM -0700, Karl Larsen (k5di at zianet.com) wrote:
> > > 
> > > sa-learn --spam --mbox /home/karl/mail/spam
> > > 
> > > 	This then drops the number of spam messages back to a couple a 
> > > day which I save for the next month.
> > 
> > Why only do this once a month?
> > 
> > I have a folder 'spam-learn' which is trained against every 30 minutes
> > (via cron).  I dump false positives in there as they appear, and clear
> > it out periodically.  Runtime (1.7GHz system) is a few seconds for a
> > dozen or so mails.  I've collected about 590 false positives over the
> > past six months (currently a few every few days).  A lot of it (~20%) is
> > Asian spam to list mail.
> 
> I've been using bogofilter in the same way.  I'd love to set up a
> sensible cron job that would unregister falsenegatives from the good
> words list and register them to the bad words list.  Like this (iirc):
> 
> bogofilter -N < /home/drew/mail/bogo/falsenegs
> bogofilter -s < /home/drew/mail/bogo/falsenegs
> 
> (I need to run the "unlearn" with -N because bogofilter auto-learns when
> it initially classifies a mail via procmail and the -u switch.)
> 
> The thing is that I can't figure out how to cleanly empty the mailbox
> with the script.  I'm sure it's easy to find, but I've been lazy.

What format's the mailbox?

With maildir, simply move from 'cur' and/or 'new' to some temporary
directory, run trainer, then delete the temporary directory.  mbox is
more problematic.


Peace.

-- 
Karsten M. Self <kmself at ix.netcom.com>        http://kmself.home.netcom.com/
 What Part of "Gestalt" don't you understand?
    George W. is deceptive to be sure. Dissembling, too. And let's not
    forget deceitful. He is lacking veracity and frankness, and void of
    sooth, though seemingly sincere in his proclivity for pretense. But
    he did not lie.
    http://www.jointhebushwhackers.com/not_a_liar.cfm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.svlug.org/archives/svlug/attachments/20031219/cc971165/attachment.bin