Just my 5 cents. I have Brightmail installed in 3 MTA machines running
proprietary software on Solaris.
Brightmail includes a SIEVE implementation for filtering out site specific
BAD email messages.
After using this, I wish every MTA would have a SIEVE filtering capability
so I don't have to learn all the different methods for filtering by the myriad
MTA software out there.
Please compare what you have written here with the subject of this thread.
Specifying filters that are applicable to more than one user, or applicable
to all users, is one thing. Specifying filters that control MTA behavior,
especially behavior as it pertains to detecting and controlling unauthorized
relay attempts, is quite another.
I view the former as being well within the purview of sieve. I view the
latter as being far out of scope.
Since we have a filtering language RFC, why not
use it? Let's simplify the world - it's complex enough.
As it happens we have not one but two drafts in this general area:
draft-degener-sieve-multiscript-00.txt and
draft-daboo-sieve-include-00.txt. So this idea is already being pursued.
But the case in point is, SIEVE is intended for final delivery filtering, not
at the MTA level.
The conflation of sieve with final delivery offers no real impediment to
performing system level actions.
Ned