The point is to design the
MTA to fit the email architecture, not to design the email
architecture to fit the current MTA design. Else you are simply
putting the cart before the horse.
We already have a large, distributed, heterogeneous, deployed
email architecture infrastructure and MTAs. Extensions need
to work with that existing, deployed infrastructure in a
backwards-compatible way. Believe it or not, the world isn't
going to switch over to completely new MTAs overnight just
because some draft says so.
I agree completely with you. That's why DMTP *is* backward-compatible
and would work with any legacy SMTP MTAs. There is no requirement for
everyone to switch to DMTP overnight at the same time. This is indeed
one of the central considerations in this proposal. The argument we
were responding to above was that it is supposedly impossible to
redesign MTA clusters -- takes some effort, but its feasible and
doesn't have to be done overnight by everybody. These systems designs
are not that static anyway, they have to be evolving even now as we
speak.
- Kartik