[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Verify that spamc is using Bayesian?

On Tue, Oct 21, 2003 at 11:02:49PM -0400, ScruLoose wrote:
> On Sat, Oct 18, 2003 at 05:52:40PM -0400, Kenneth Dombrowski wrote:
> > On 03-10-18 12:47 -0400, ScruLoose wrote:
> > > 
> <SNIP> 
> So if I'm reading things right, the situation is like so:
> 1)  If I want to smtp-time reject mail based on SA's analysis, I need to
> have SA called systemwide as seen here
> http://dman13.dyndns.org/~dman/config_docs/exim-spamassassin/node14.html
> Which forces me to have one system-wide Bayes database.
> 2)  If I want to have per-user Bayes databases, I must install procmail,
> and call [spamassassin|spamc] from ~/.procmailrc.
> Exim's .forward filtering (for some reason) cannot do this.
> 3)  If I want both smtp-time reject for the really obvious spam _and_
> per-user bayes DBs for the rest, I'd have to integrate SA-exim
> systemwide, configured _not_ to use bayesian tests, then call SA _again_
> (or bogofilter or somesuch) from each user's .procmailrc (_not_ an exim
> filter .forward).

Maybe don't train a system-wide SA. Without training it should not do
Bayes. Or maybe adjust its threshold for rejection setting. 

> Can anyone confirm or deny?

Paul E Condon           

Reply to: