The significance of word placement

[David Relson]

At 09:36 AM 10/25/02, Boris 'pi' Piwinger wrote:

>Matthias Andree wrote:
>
> >> So do, some don't. Combined it is not enough. This is really
> >> every day language, so it is hard to know for bogofilter.
> >> Just looking at the subject, it is trivial for procmail.
> >
> > So go the Unix way: have procmail filter with static rules like
> > "Subject: Betreff" and let bogofilter deal with the rest.
>
>My hope was to remove all those rules in favor of bogofilter.
>
>pi

pi,

It's nice when 1 tool can do everything.  However, bogofilter's approach 
may not fit that particular need.

A couple of weeks ago I saw/got your procmail rule for handling asian 
languages.  Adding that to /etc/procmailrc pleased me immensely and now I'm 
thinking about the best way to remove the no-longer needed "asian tokens" 
from my wordlists.  To my way of thinking using two tools here (procmail 
and bogofilter) is much better.

You'll laugh at my old way of dealing with spam.  I saved it all, ran it 
through a program to extract the source ip addresses, then blocked smtp 
connections from _all_ of those addresses.  Given that the list was 2000 
and growing, this wasn't very efficient.  Also, it was blind to spam from 
new (different) addresses.  Bogofilter is working much, much, much better.

David