Eliot Lear allegedly wrote on 07/30/2009 08:41 GMT+02:00:
On 7/30/09 8:21 AM, Tony Li wrote:
All of the old ones have great difficulty in changing the L3 point of
attachment, and end up using the Home Agent concept as the rendezvous
point. If location is independent of identity, then it's conceivable
that one can update peer locations without loosing connections.
What about HIP?
HIP doesn't handle both ends moving without at least one rendezvous
server. I don't know of anything that doesn't need some kind of stable
point.
Tony Li allegedly wrote on 07/30/2009 10:15 GMT+02:00:
Yes, ok, if you go to the trouble of CGI's then you can change too.
Wouldn't it just be simpler and cleaner to have location be separate?
(1) it's possible to run HIP with only control messages secured, or
various levels of security on the data packets. (2) HIP is getting put
in chips ...
... but (3) I don't understand, how does HIP not separate location?
Darrel Lewis (darlewis) allegedly wrote on 07/30/2009 10:19 GMT+02:00:
With HIP, how do two correspondents (which negotiate a HITs between
them) refer to a 3rd which they don't communicate with at the moment?
Ephemeral IDs are problematic in the real world, since global naming and
referrals today tends to use unique and semi-permanent IDs.
Referrals are another problem. In a simple world you just go look up
the referred HI or address or other stable identifier in a registry,
like DNS or whatever you used to find the first peer you're talking to.
In a complex world, you may need something like
http://tools.ietf.org/html/draft-carpenter-behave-referral-object (which
was presented in BEHAVE).
swb
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
