On Jul 15, 2009, at 10:52 AM, Scott Brim wrote:
Lixia Zhang allegedly wrote on 07/15/2009 1:42 PM:
Do you think the session need to have a globally unique identifier
or
locally agreed between the endpoints during the contact
establishment?
I've seen multiple mentioning about "sessions", but yet to see a
clear
definition of it.
Lixia,
First informally: You know what it is. Why do you need it rigorously
defined?
1/ I honestly do not know. Citing Joel's comment here: does a session
mean the combination of two ends? do we care unicast/multicast
sessions? how lasting/transient is a session (hence the life time of
its ID)? etc etc.
2/ more fundamentally: when I program, maybe I could say "I know what
each part does" and dont define rigorous interface (but still, that is
NOT good practice)
But here we are talking about a big architecture. As I've seen, even a
term with a definition often gets different interpretation by
different people, let alone a term without a clear definition!
But I would like to step up a level and repeat what I said earlier:
1/ I do not think it is in our charter to define how many identifiers
we ought to have, or what they ought to be.
2/ Our job is to figureout scalable routing architecture.
3/ We need to have a good understanding about the interplay between
addresses and identifiers, no less and also no more.
Related to the last point: I had planned to get a short draft out on
the terminology discussion we had at SF IETF before the draft 00
deadline but you all know by now that I did not get there. I am still
trying to finish it in coming days, to get it ready for discussion at
Stockholm.
Lixia
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
