ietf-822
[Top] [All Lists]

Re: MIME types for ack[-request]?

1992-03-26 15:36:24
Nathaniel,

Certainly, Mime offers an adequate base for acks, etc.  My hope is that
we can add a proper framework for control messages, rather than
doing things entirely piecemeal.

Personally, I expect read-ack to be the most useful, too, tho I do
hope that delivery-acks could be made to work, since I would dearly
love to have end-to-end acknowledgements of delivery.

My suggestion is that a major content type of "Control" be added,
with subtypes of "Sender-MHS", "MHS-MHS", "MHS-Receiver", and
"Sender-Receiver".  (I've no religion about the vocabulary, per se.)

If we really view these as protocols, then we need to capture the
Request/Reply status and this could be done either through
the major type (Control-Request vs. Control-Reply) or the subtype
names (e.g., Sender-MHS-Request vs Sender-MHS-Reply), or as a parameter
of the subtype.  SNMP uses the middle style.  While it creates 8
subtypes, I lean towards it, somewhat.  

Hence, your own mechanism becomes "Control/Sender-Receiver-Request:
action=read-ack" which returns a "Control/Sender-Receiver-Reply: 
action=read-ack".

Please note, I'm suggesting we take a bit of time and develop a clean,
extensible framework, but not a particularly elaborate one.  Then we
can a) think quite a bit more clearly about specific functions, and
b) add them cleanly.  Stuffing everything into a type of Application
strikes me as not scaling very well.

Dave


P.S.  Wicked thought:  Would it be useful to have application/snmp?

d/