spf-discuss
[Top] [All Lists]

RE: Dealing with SPF problems

2005-04-05 04:55:00
...... Original Message .......
On Tue, 5 Apr 2005 09:31:06 +0200 (CEST) "Arjen de Korte" 
<spf+discuss(_at_)de-korte(_dot_)org> wrote:

But please don't give a bad score for NEUTRAL.  It's supposed to be just
like no SPF at all.

Not at all. In a scoring system (SpamAssassin in this case), anything goes
as long as it says something about the likelyhood of a message being spam
or non-spam. Since I receive about 100 times more spam than non-spam from
some US based ISP's, I award them a small additional score. This probably
won't work for many people, but in my setup it does help weeding out spam
from the non-spam.

If you find that a disproportional amount of spam is from SPF NEUTRAL
systems, it is valid to use this and award points for it. I have no doubt
that in many cases the Baysian filter will pick up the difference
automatically. You are right that one shouldn't outright block messages
with SPF 'neutral' or 'softfail', but there is no reason why you couldn't
use them in a scoring system. That's why I submitted this patch to
SpamAssassin.

Best regards,
Arjen
-- 

I understand that, but for many domains on a shared MTA, NEUTRAL is the 
best one should do.  Treating NEUTRAL that matches a mechanism and NEUTRAL 
from ?all the same combines two different things into one.  

From a SA statistical perspective you are right.  I don't think that it's 
good for SPF.

Scott Kitterman


<Prev in Thread] Current Thread [Next in Thread>