Quinn Comendant wrote:
> It's not necessary, but it's a pretty darned good idea.
>
> Are you running any periodical (cron) sync or expire options with sa-learn?
Not yet. I've been leaving that out in hopes of pinpointing what the problem
really is. I wait for it to break, run something, and wait for it to break
again. Once I'm confident with what's going on, I'll add a daily cron job.
> Training mechanisms?
Not yet.
> I'm venturing to setup a setup within Horde/IMP so users can flag spam/ham
> message to help train the database. But I'm still a bit short in the know-how
> of executing scripts via valias (see my related thread).
Sounds good.
> Configuring and tuning SA is a top priority for the QT in my opinion.
I agree.
> Q
>
I'm not quite so eager to set up a whole training mechanism yet. From a
philosophical point of view, I still believe that bayesian type of filtering
is best done at the user level. SA only does this globally, and if you're
running multiple domains the problem is even worse.
That's why I'm interested in seeing DSpam implemented in the toaster. I've
heard of some servers running both SA and DSpam (in that order). I'd like to
see how well DSpam does on its own first, then maybe use SA with bayes
turned off along with DSpam.
I'm thinking it'll be several months though before we get DSpam running on a
toaster. In the meantime, I'm simply try to get SA stabilized.
--
-Eric 'shubes'
---------------------------------------------------------------------
QmailToaster hosted by: VR Hosted <http://www.vr.org>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
