On Wed, Nov 09, 2005 at 10:44:18AM +0100, Tom Kistner wrote:
> Michael Haardt wrote:
>
> >> deny message = Spam detected
> >> !authenticated = *
> >> spam = incoming-$primary_hostname/servers=spamd1 783 : \
> >> spamd2 783 : ...
>
> >run multiple *different* spamd servers since some years now. But the
> >result of the discussion was: Unless it stays compatible with the awkward
> >syntax, there will be no change.
>
> I consider the "spam" line quoted above to be awkward. :)
If you look at the specific example, I agree. But being able to specify
the spamd server in the ACL is way better than the current approach.
If I got the original poster right, he wants to run two independent SA
servers with different configuration. I run two independent SA-compatible
servers, both being entirely different pieces of software.
> I'm with Tony here: You already can send a username to spamd. Changes in
> rule treatment should be implemented in spamd, using the submitted
> username as the key. If spamd can't currently do this, hacking on spamd
> instead of exim is the better choice.
It may solve a few problems, but certainly not all. I think we had that
discussion already.
> Anyhow: If anyone submits a patch that adds the above "servers" option
> to the "spam" condition, i'll add it. It should completely override the
> global spamd_address option when present.
Adding expansion? I wouldn't want to double a rule, using a condition
to select one or the other, just to avoid expanding the "spam" condition.
> >+ /* if spamd_address starts with a dollar, expand it first */
> >+ if (*spamd_address == '$') {
>
> Now thats REALLY horrible :)
In my opinion, horrible is even too nice a word for that patch, but:
/* if av_scanner starts with a dollar, expand it first */
if (*av_scanner == '$') {
I just followed the style, hoping you would accept this in order to
finally reduce the amount of private patches I carry from release
to release. :)
Michael
