Administrateur Systeme sysadmin at DMS.UMONTREAL.CA
Thu Nov 14 15:45:31 GMT 2002

Tim Sellar wrote:

>There is a relevant artivle showing Bogofilter vs. SpamAssassin  here:
>Given bogofilters huge performance advantage I am considering implementing
>it as a primary spam filter. Mail which gets past it would then be passed to
>Spam Assassin acting as a secondary filter. I would hope this gives the best
>of both worlds. As a system wide filter I cannot guarantee the effectiveness
>of any initial training I apply to bogofilter to reflect an individuals
>particular email but can hopefully rely on Spam Assassin to catch most of
>what bogofilter misses. Over time, spam trapped by SA will be fed back
>(automatically?) to improve bogofilter's effectiveness and reduce reliance
>on SA - hopefully giving better system performance...
>Anyone see any flaws with this two-pronged approach?
>What this does mean is I need to look at getting Mailscanner working with
>both Exim and bogofilter...
Doesn't the SA 2.50 incorporate a bayesian filter along the lines of
P. Graham's paper ?


                     Christopher Albert
            Responsable des services informatiques
         Departement de mathematiques et de statistique
                  Universite de Montreal

           bureau 6188, Pavillon Andre-Aisenstadt
          Tel: (514) 343-2281  Fax: (514) 343-5700

More information about the MailScanner mailing list