[Gllug] Newly installed SpamAssassin less effective than an old installation (same version)

John Winters john at sinodun.org.uk
Sat Jan 14 11:12:55 UTC 2006


On Sat, 2006-01-14 at 10:58 +0000, Nix wrote:
> On Sat, 14 Jan 2006, Martin A. Brooks prattled cheerily:
> > John Winters wrote:
> >> I have not customised SA on either machine.
> > 
> > Did you have Bayesian learning switched on?
> 
> Do you have trusted_networks set?
> 
> (If not, please set it properly ;) )

I have just a default Sarge installation of SA.  Can you give me some
pointers on where to find out what "trusted_networks" means?

> 
> > If so the new machine probably hasn't yet seen enough messages to
> > start adding bayes weighting to message scores.
> 
> I'd not expect that to push hit rates down to only 10%; that's far lower
> even than the -set0 (no Bayes, no net tests) should score.

Me neither.

> What sort of rules get hit? Could you post an example of something that
> SA misses which you'd guess it shouldn't?

Tricky, since if I forward such a message to this list it will probably
get binned by all your spam filters.  However, I'll give it a go.  Spam
message closely follows this one.

John

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list