Daryl C. W. O'Shea wrote:
Mat's not on this list, so CC him if you reply.
(forwarded to [email protected])
From: Matthew Sullivan <[EMAIL PROTECTED]>
I basically suggest Spamhaus should now be placed in the same config for
SA as MAPS, default off (score 0) as it has caused me quite some trouble.
We started to discuss this a couple weeks ago.
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5641
Is there anything I can do to help?
Consider making the default not to use Spamhaus, and therefore base your
regression tests without Spamhaus, because 3.2.x is nowhere near as good
as 3.1.x for identifying spam without false positives.
I'm not sure that the FP rate is directly caused by the Spamhaus rules.
If it is it could be that one or more or the limited number of corpus
submitters have adjusted their mail infrastructure to avoid Spamhaus FPs
thus artificially allowing the rules to be scored higher than they
should have been. I know I have re-designed my mail layout in order to
avoid FPs on SORBS DUHL. It's far more likely that our ham corpora are
just not diverse enough, though, if you're seeing a significantly higher
number of Spamhaus FPs than we are.
Daryl
