Re: [Isms] Transport protocol discussion

[Eric Rescorla <ekr at rtfm.com>]

"Tom Petch" <nwnetworks at dial.pipex.com> writes:

>> My $.02:
>>
>> The important transport question is whether support for datagram
>> mode is a requirement. If it is, that substantially constrains
>> the solution set--and potentially requires some more design work :)
>>
>> -Ekr
>>
>
> You've lost me here and I would like to understand.
>
> I see SNMP applications as datagram, message out message(s) back, not stream,
> not full duplex.  If we have a full duplex stream transport, then a SNMP-like
> message exchange can be imposed on it, as the SNMP over TCP RFC demonstrates;
> needs work, but can be done.

Well, that's not datagram, that's record-oriented stream transport.
The key difference being reliability and in-sequence delivery.

One scenario where this is important is in real-time multimedia
applications. If a packet gets lost, having it get retransmitted
500ms or 1s later doesn't do much good. What you *do* want is
the next packet. But in reliable stream transport like TCP,
packet n+1 can't be read until packet n is received.

I'm not enough of an SNMP expert to say whether this is important
in SNMP....

-Ekr

_______________________________________________
Isms mailing list
Isms at lists.ietf.org
https://www1.ietf.org/mailman/listinfo/isms