Glenn Anderson wrote:
At 10:55 am -0400 10/8/2008, Hector Santos wrote:
>
Ok, lets suppose that on your 2nd try with USER2, that it now ACCEPTED.
Please explain why did user1 now lose out in getting the mail when it
was a more valid 250 case and user2 was not?
That would be entirely up to the receiving server. I am certainly not
going to assume that there are no cases where this would happen, that
could result in lost mail.
Thank you, and now we come in full circle of the original problem Tony
Finch highlighted.
The thing is, the pattern provided here is very common, and in this
regard, the proper behavior is to follow the specs. It's correct.
Unfortunately, we had 8 years of a incorrect semantic in 2821 that
fell through the cracks. I hope that implementators noticed it.
If the server wanted to deliver to USER1, it must issue DATA 250.
USER2 and USER2 was not expected to get "that" message.
However, this will allow you to properly continue with retries for
USER2 for "that" message. That is the scenario Tony Finch highlighted
that lacks proper semantics in the specs.
And IMO, this will help match the single transaction for recipient
method that I think everyone wants.
--
Sincerely
Hector Santos, CTO
http://www.santronics.com
http://santronics.blogspot.com