Keith Moore (moore@cs.utk.edu)
Wed, 12 Jan 2000 08:47:00 -0500
> However, we could have a sending system
> that parses the DSN. However, it doesn't understand the difference between
> an RFC 1894 DSN (whole message failed) and a PNDN (part of the message
> succeeded). In this case, aren't we in the same situation as a system that
> doesn't understand PNDN?
For a sender that understands DSN but doesn't understand PNDN,
I think the choice is one of the following:
- sender understands that there was some sort of message delivery problem,
and perhaps chooses to resend the entire message.
- sender doesn't understand that there was a message delivery problem
because he/she/it doesn't know how to parse the PNDN at all.
To me, the first choice sounds like the safer and more conservative one.
In general, anytime a delivery agent cannot deliver a message with
complete fidelity it has a choice - should this be reported as
"partial delivery" (thus a form of success) or delivery failure?
There are lots of potential cases that could be labelled partial
delivery - such as when the main portion of the message can be
delivered but some attachments cannot, or when all parts
can be transferred but there is likely to be some conversion loss
(say when a color image gets gatewayed via black-and-white fax),
or when some but not all pages of a fax can be delivered.
The choice of whether to label such an event (partial) success or
failure seems like one which should be very dependent on the
specific circumstances of the the message and its delivery or
conversion. Of course it would help to have some guidance from
the standards on when to label something a success and when
to label it a failure.
We grappled with this to some degree while working on DSNs,
which is why there are status codes for both
'conversion required but not supported' (presumably a failure)
and 'conversion with loss performed' (presumably a success).
Also, I think it would be odd for us to use DSN to report
delivery failure, and also to use DSN to report complete delivery
success, but to use some other format for reporting partial success.
Seems like it is cleaner to use DSN for everything, to choose
a 2.x.x. or 4.x.x or 5.x.x status code as appropriate for
the overall message (new status codes could be defined for fax),
and to add additional fields to the DSN to provide more detail
about which parts of the message were undeliverable.
Keith
This archive was generated by hypermail 2.0b3 on Wed Jan 12 2000 - 15:50:03 IST