Eric,
EAH> On 5/29/2004 9:28 AM, Dave Crocker wrote:
Certainly the role of independent mta-based routing needs to be cited.
I'm not so sure that re-queuing does. I think of email's queuing as a
(popular) local implementation issue.
EAH> 4xx means something completely different at:
Well, the 4yz response code model is actually the only reason I think
queuing _might_ be in the architecture.
EAH> The peculiarities are dictated in the specs accordingly. More to my point
EAH> however, is that the [re-]queuing algorithms in the specs also gives
EAH> instructions on what to do in each of those cases.
Well...
2821> 4.2.1 Reply Code Severities and Theory
...
2821> ... but the SMTP
2821> client is encouraged to try again. A rule of thumb to determine
and
2821> 4.2.5 Reply Codes After DATA and the Subsequent <CRLF>.<CRLF>
2821> When an SMTP server returns a positive completion status (2yz code)
2821> after the DATA command is completed with <CRLF>.<CRLF>, it accepts
2821> responsibility for:
...
2821> - if attempts to deliver the message fail due to transient
2821> conditions, retrying delivery some reasonable number of times at
2821> intervals as specified in section 4.5.4.
are examples of the directives given in the specification. By any
reasonable measure of protocol specification, they are pretty darn weak.
Still, the text "...an SMTP server... it accepts responsibility for...
retrying delivery" would seem, to me, to eliminate any question. It is
an explicit statement that retrying is part of the model.
d/
--
Dave Crocker <mailto:dcrocker(_at_)brandenburg(_dot_)com>
Brandenburg InternetWorking <http://www.brandenburg.com>
Sunnyvale, CA USA <tel:+1.408.246.8253>, <fax:+1.866.358.5301>