On Wed, Mar 06, 2002 at 08:06:55PM +0000, Philip Hazel wrote:
> On Wed, 6 Mar 2002, Marc MERLIN wrote:
>
> > Picture the requirement "sales@domain has to get *everything*" If that means
> > that a mail send to sales and some other person and that fails header
> > callbacks goes through to both, no big deal.
>
> Seems a messy specification to me. "Other person" may well complain ("I
> specifically asked not to get this crap.")
Let's put it a different way.
CEO says: "we will not bounce any mail to sales@companyname"
Options are:
- turn off every single exim check (argh....)
- implement something that will insure that checks are disabled for some
receipients.
The fact that some other person may get a spam is sent to sales@ and Cce to
him (which in real life should pretty much never happen) is irrelevant, and
I'll be glad to talk to him/her if (s)he feels this is a problem :-)
I can assure you that in the corporate world, bouncing some mail no matter
the technical reasons, can be unacceptable.
This means that solutions which reduce bad Emails but allow a few ones here
and there become very desirable.
> This is beginning to sound like the kind of special feature that
> somebody wants but which I can't see as being cleanly implementable. For
> that purpose I invented the local_scan() function. You can write your
> own logic in C and use it to tell Exim whether to say "yes" or "no" at
> DATA time. Admittedly, in your case, you'll have to check the headers
> yourself - there isn't an interface to Exim's checking code.
Mmmh.
I suppose I could run local_scan first, and if it sees a local_part it
likes, it just returns accept.
Otherwise, header_check and header_verify get to run.
I'll see which one requires the least work between this and hacking exim
directly (I'll have to see whether I can access a the list of receipients
for a mail when I'm in the DATA ACL)
Marc
--
Microsoft is to operating systems & security ....
.... what McDonalds is to gourmet cooking
Home page:
http://marc.merlins.org/ | Finger marc_f@??? for PGP key
