Hello Ted,

On Wed, 08.12.2004 at 14:43:17 -0500, Ted Zlatanov <[EMAIL PROTECTED]> wrote:
> > I use qmail-qfilter to deliver spam-filtered mail, like this:
> > 
> > #!/bin/bash
> > 
> > exec /var/qmail/bin/qmail-qfilter \
> >     /opt/csw/bin/spamc -- \
> >     /var/qmail/bin/new-inject -n
> > 
> > I would like to give the "-u USERNAME" parameter to spamc so it can
> > use per-user configurations, but that information is not in the
> > environment (I checked with `env`) and the script above is run by user
> > qmaild.  When qmail-local is called the information is in the
> 
> To clarify: I understand that the RCPT is not the same as the final
> delivery address.  Would it be better, in any case, to use spamc as a
> delivery filter in the last stage of delivery?  I would prefer to
> avoid running it N times for N users if the mail is going to a mailing
> list I host but if I must then I must.

you probably have an error in your logic. You usually only have one of
these:

- one SA run per recipient, thus obeying per-user settings, or

- one global SA run per message, thus ignoring per-user settings.


Except in case you find it useful to run SA both in front of qmail-send
and behind, too, first implementing some global policies, then,
afterwards, adding per-user adjustments. If you want to exclude some
mailing lists from running SA on every recipient (where it probably
makes the least sense anyway), you will probably find yourself running
SA both in frond and behind qmail-send, and some trickery in the
invocation scripts to select the stage at which you run for each
message individually.


Best,
--Toni++

Reply via email to