To some extent, these questions are orthogonal to the questions about a
potential relationship between HIP and P2PSIP.
Using HIP in a decentralized manner requires a distributed rendezvous
service (and distributed name service). An overlay such as have been
proposed by the various peer protocol proposals is ideal for running
such services.
P2PSIP requires a distributed registrar service. This service also
requires a peer protocol (and some layers on top) and is not supplied by
HIP.
Now there are still some architectural questions of whether the peer
protocol connections should be formed using HIP or not and also
questions about how to provide the best interface for applications using
the services.
Regardless of the answers to those questions, I hope we can develop a
peer protocol that can be extended to support the various models for
interacting with HIP (and work without HIP) and let future development
and deployments determine what architectures are truly useful.
Bruce
Dean Willis wrote:
On Jun 26, 2008, at 1:42 AM, Gonzalo Camarillo wrote:
Hi Henry,
Yes, I know, developing HIP code looks like opening a whole new can of
worms, but nothing compares to what we are looking at now when trying to
traverse NAT, support mobility, multihoming, etc. for each application
protocol and their various flavors separately.
yes, that is what HIP is about (i.e., implementing those functions at
a lower layer so that they do not have to be redesigned by every
single application-layer protocol).
This asks the question "Why don't we believe in HIP in this role?"
Is it because we've seen HIP struggling to advance for many years and
think we can move more quickly?
Is it because we think the IETF's immune system will suppress HIP but
that application-level work can move through?
Is it because we think that doing this stuff at the HIP level requires
widespread OS and IP stack changes, but that we can deploy
application-level solutions without it?
Is it because we think that if HIP solves the problems, then there will
be no fun work left to do on applications?
Or is there something else?
There must be some reason, as I would think that if people really
believed in HIP that the entire resources of the IETF would be bent
towards getting it wrapped up and ready to go, since solving these
problems again and again for every different application makes no more
sense than would reinventing IP for every application.
--
Dean
_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip
_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip
