[Gllug] RE: Gllug Digest, Vol 19, Issue 75

[Nix]

On Mon, 31 Jan 2005, David Abbishaw stipulated:
> I have been bouncing it back as unknown, unfortunately this seems to clog up
> the queue and the badmail folder and so far doesn't appear to have reduced
> the incoming mail that much.  For anything that's on the SPF list I don't
> bother generating an NDR as it wont get there.

As others have pointed out, reject, don't bounce: it avoids clogging up
your mqueue. It doesn't necessarily stop your upstream relay (if any)
generating a DSN itself, though.

> Why do you want the spam?

I am (or rather, will, when I've written some scriptage to identify and
eliminate multiple substantively identical copies of the same spam)
submitting mass-check results to the SpamAssassin project, for use by
the perceptron that fixes optimal score values.

The perceptron is actually so fast that I might run it myself on only my
own training set, but I'm worried about overfitting to the data set
(we've historically had problems with that) so the combination of mine
and everyone else's ham and spam might help reduce the extent of the
problem by at least overfitting to a training set that's not *entirely*
mine. :)

-- 
`Blish is clearly in love with language. Unfortunately,
 language dislikes him intensely.' --- Russ Allbery
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug