On 2/15/2007 4:04 AM, Tony Finch wrote:
On Wed, 14 Feb 2007, Eric A. Hall wrote:
If the write fails then the client wouldn't have received it.
And the server is supposed to drop the message *after* it has committed to
handling it?
That's what duplicate avoidance means--dropping duplicates.
But maybe we are making more of an issue out of this than it really is.
If duplicate avoidance really is "a problem" and not an annoyance, then
some kind of checkpoint mechanism with deliberate client acknowledgments
to the server acknowledgment would be the way to go. It's heavy but
appropriate for something that is really and truly a problem that needs to
be solved.
If it's just an annoyance, then something as simple as "detect errors" as
suggested is probably overkill.
Either way, it's a problem for SMTP and not a problem for this extension,
so it ought to be dealt with separately.
--
Eric A. Hall http://www.ehsco.com/
Internet Core Protocols http://www.oreilly.com/catalog/coreprot/