[OT] Spamassassin marking list messages as spam (was: debianuser@brown.edu)

[Todd]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Mark Kirchner wrote:
> On Thursday, January 22, 2004, 3:06:33 PM, Scott wrote:
>> The rules for filtering spam were probably written by someone who thinks
>> Debian is an 'all natural' V-!- at -G-R-@ substitute.   
>
> Hmm, I don't think so. :-)

Nor do I.

> The X-Spam-Status Line says:
> | X-Spam-Status: Yes, hits=6.2 required=5.0 tests=RCVD_IN_MULTIPLE_RBL,
> |         RCVD_IN_NJABL,RCVD_IN_SORBS autolearn=no version=2.62
>
> The successful tests (RCVD_IN_MULTIPLE_RBL, RCVD_IN_NJABL,
> RCVD_IN_SORBS) seem to indicate that the sending host (or some other
> host host in the "Received:"-lines of the mail) has entries in several
> (anti-spam-)blacklists.
> Well, at least, I hope I'm not totally off with this explanation...

I think you're right on here Mark.  I was just getting ready to send
something similar.

I run Spamassassin locally, with Bayesian filtering and non of these
messages have been marked as spam.  That's thanks to the extra check
of the Bayesian filter AFAICT.  Here's my SA output:

X-Spam-Status: No, hits=-4.7 required=5.0 tests=BAYES_00,RCVD_IN_NJABL,
        RCVD_IN_SORBS autolearn=no version=2.61

The main difference is the BAYES_00 test, which should lower the score
a good bit.

The mail is most definitely not spam.  Jean-David, you ought to take
whatever action your ISP provides to have that message marked as ham
(non-spam) so they can learn from it and not mark similar messages as
spam in the future.

I've found, using SA locally with Bayesian filtering and a decent
sample of spam and ham to train it with, SA has a near zero level of
false positives.  I've had one in about 6 months and it was one that
mocked the Nigerian money scams.  Spamassassin is an incredible tool,
IMO.

- -- 
Todd        OpenPGP -> KeyID: 0xD654075A | URL: www.pobox.com/~tmz/pgp
======================================================================
What do Atheists scream during sex?
'Ohhh, chemical chance, chemical chance'
    -- Bill Hicks

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: When crypto is outlawed bayl bhgynjf jvyy unir cevinpl.

iD8DBQFAD/YEuv+09NZUB1oRAl5jAJ4mBl9uRoWx3YTdyo207PgnIbPG1wCg+tCb
PBLcz0cAkTjAP6uIrEvSU5E=
=SXRc
-----END PGP SIGNATURE-----