Migrating from SpamAssassin to Bogofilter

RW rwmaillists at googlemail.com
Thu Jun 6 15:34:05 CEST 2013


On Wed, 05 Jun 2013 13:56:33 -0700
Kip Warner wrote:

> On Wed, 2013-06-05 at 08:43 +0200, Matthias Andree wrote:
> > Greetings,


> > Note that bogofilter is a purely Bayesian classifier, and does not
> > do any of the other checks SpamAssassin does.
> 
> Yes, I understand. But given that my SpamAssassin installation is so
> broken right now and upstream hasn't been much help, I really don't
> have any choice but to switch.

I don't want anyone to get the wrong impression about that. When you
say "upstream hasn't been much help", you demonstrated one certain bug
in the Evolution plugin and evidence of further plugin problems. In
every test performed outside the plugin, spamc/spamd behaved correctly.
SA developers can't be expected to debug Evolution.

 
> > (I also tried spamassassin's Bayesian mode a few times and found it
> > to be awkwardly slow -- not sure if that has changed.)
> 
> I don't know what mode I was using, but I do know that it was pretty
> slow.

The default configuration has network tests enabled by default (and I
don't recommend turning them off); in this case performance is typically
dominated by waiting for the slower remote servers. Bayes and other
local tests are performed whilst waiting for responses. In the past
I've seen SA with pre-cached network tests and with Bayes enabled,
average 30+ emails per second on a low-end desktop from a few years
back, so I wouldn't say Bayes is all that slow.

Bayes can be a bottleneck on busy servers, but the defaults are
optimized for ease of use, not speed.



More information about the Bogofilter mailing list