Hi, I'm having some problems with retraining messages marked
incorrectly as spam or notspam with postfix. The issue seems to be
that the recipient of the mail is a postfix alias (or virtual map,
doesn't seem to matter) that is directed to a real user on the
system.

The problem arises in that if a user has multiple aliases, and spam
is arriving on one of them that they need to retrain, if they bounce
it to their "real username"-spam, dspam doesn't find the signature
and gives the "Unable to find a valid signature" error. If they
bounce it to the alias that the spam was sent to, all seems to work.

I direct mail through dspam using
http://dspamwiki.expass.de/Installation/Postfix/NealesSetup for the
most part (ie, direct mail to dspam via a pcre user access filter).

However, instead of using his perl script for retraining, I use a
couple of transport aliases:

/-spam\@(.*)$/ dspam-retrain:spam
/-notspam\@(.*)$/ dspam-retrain:innocent

and a master.cf rule:

dspam-retrain unix - n n - - pipe                                    
                flags=Rhq user=dspam argv=/usr/local/bin/dspam \     
                          --user $user --class=$nexthop
--source=error

I've seen a couple of posts on the list about this, the most
promising one seemed to be to recompile with --enable-virtual-users.
However this does not seem to have fixed my problem.

The perl script mentioned by Neal probably would fix the problem if I
parsed for the original recipient, but then I have to augment my
selinux policy a lot or pretty much disable selinux to allow enough
functionality for postfix to execute this script.. I really would
like to avoid this. Is there any other option?

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

Reply via email to