> Well, header I have on detected spams like these (possibly I need to
> reconfigure something) to get the above lists:
>
> X-Spam-Status: No, score=-1.8 required=1.5 tests=BAYES_00,
> DATE_IN_FUTURE_03_06 autolearn=ham version=3.0.4
This says that the message is NOT spam. Bayes in particular is convinced
that it was ham, and has auto-learned it as such. It looks like your Bayes
database is corrupt and will need to be rebuilt at some point.
However, before doing that, we need to figure out why all this hit was
date-in-future.
I'm a little concerned that that initial Received header from Qmail may be
keeping SA from analyzing the return headers. This could case a lot fo the
net tests to fail. Hiwever, maybe that header is ok. Others that know more
about received headers than I do will look at it.
If that spam had been in English the body text would have nailed it.
"Greencard lottery" is pretty much a guaranteed spam indicator all by
itself. This looks like a case where a few of the rules for these things
could profitably be translated to German, by someone that speaks both
languages. (I don't, or I'd just do it.)
There is a good chance though that register-usa.com should have hit SURBL.
This makes me think your net tests aren't working. Since this is 3.0.4,
they should be by default.
Why don't you run "spamassassin --lint -D" and post the output. My guess is
you either have a problem with the Net::DNS version, or init.pre insn't
installed correctly. (Of course you might be starting SA with the -L
parameter that disables net tests.)
Loren
