John C Klensin wrote:
I don't disagree with anything Ned said in his note, but want to
reinforce the principle that trying to make too-fine
distinctions in this area won't work in practice and may lead to
even more confusion.
I agree, that is why we need to stick with a long time burn-in stone
*baseline* for a state machine client/server protocol model that is
driven by reply codes.
We can't afford subjective "Consensus by Osmosis" creativity to alter
the principle meanings of the state machine.
In lieu of SMTP v3.0 having considerations for multiple reply codes
per recipient DATA responses, the current state machine model DATA 554
does not means that some RCPT will still get the mail and that others
that did not could retry the same transaction at some later point.
That is just fundamental broken.
Thats not the say that a system could ignore it and do what it wants,
but it does say that we need to stick with a baseline principle a 250
means acceptance. 45z/55z does not.
Otherwise, whats the point?
We might as well tell clients to always continue trying and then lets
come up with a new code 999 that says
"You can continue trying until you see DATA 999. When you see 999.
STOP!!"
Thats what I am basically hearing.
At what point does the transaction and retries ends?
I say the basic principle has been laid down since 821, acceptance
means a (250) OK response. Anything else is not acceptance.
--
Sincerely
Hector Santos, CTO
http://www.santronics.com
http://santronics.blogspot.com