>...
>Matt Kettler wrote on Mon, 12 Dec 2005 16:13:21 -0500:
>
>> Others would say they trust it explicitly and would
>> immediately give it 10.0.
>
>If I trust it I use it at MTA level. My opinion ;-)
>
>Kai
>
>--
>Kai Schätzl, Berlin, Germany
>Get your web at Conactive Internet Services: http://www.conactive.com
>
Kai,
I don't know about that. I "trust" all the RFCI lists, but was
blasted (Hi Matt) for using "postmaster" at the MTA level (and don't any
more). The issue is not all lists are intended to be "spam" lists, even
though they may have a high correlation with spam content. And I'm sure
that I'm still stricter than > 95% of the people on this list (though you've
mentioned local rules that would cause me problems, with my own domain and
others - e.g. a single '.' FQDN rule for HELO/EHLO - my case is a "corner"
case at best, but look at ibm.com, sgi.com and microsoft.com - who all do
or have sent mail with the domain as the HELO/EHLO argument, but also all
have 'A' RRs for them - I don't).
Also, for anyone who does have access to the MAPS lists, along with
the RCVD_* (or MTA level) rules, also consider using a DSN_FROM_* rule for
the "RBL" (that is the only one for which it is likely appropriate). Also
a URI_* rule on the "RBL" may be helpful: Neither of these additional rules
is easily done at the MTA level.
I do generally agree about the RCVD_IN_* rules *usually* being both
cheaper and more effective to use at the MTA level (but not all sites can
afford even low FP rates I can, and I guess that you can also). Still, there
will always be some exceptions that the administrator may have to work around.
Also I use meta-rules with RCVD_IN_* from lists that I *don't* trust,
but I find that the meta-rules do have a low FP rate. If 4 independent rules
each have a 30% FP rate, a meta-rule combining them in "AND" format has only a
0.81% FP rate (of course, if they are not independent, get out your statistics
texts). "Chinese" menu constructions (one from column A...) of high FP rate
rules can be used to form quite powerful meta rules with low FP rates (and
with the advantage of very low computational overhead and memory use - some
of the readers here have heard this from me before, but not in "public" or
"on-list").
Paul Shupak
[EMAIL PROTECTED]
