Dan York wrote:
Eric,
Sorry, I should have been clearer in my statement:
------
*When used with HIP*, it seems like P2PSIP needs to sit on top of HIP for the
addressing HIP provides....
------
I do realize that currently-proposed P2PSIP protocols will work fine without
HIP.
Thanks for the info below,
Dan
Let's be a bit careful about what "when used with hip" means. There are
at least four cases:
a) P2PSIP overlay deployed in a pre-existing HIP environment
b) P2PSIP overlay deployed on a traditional IP network with applications
using a pre-existing HIP environment
c) P2PSIP overlay providing rendezvous and lookup services for a HIP
environment (presumably for other applications), but not using HIP for
its connections
d) P2PSIP overlay providing rendezvous and lookup services for a HIP
environment and using HIP for its connections
Case (a) works with any peer protocol that has been proposed since the
peer protocol doesn't know it's using HIP. In this event you could even
deploy the peer protocol without ICE since HIP is doing the ICE
negotiation for you. (of course, you need to know this)
In case (b), if the resources stored in the overlay use HITs, neither
the peer protocol nor the application need to be aware HIP is being used.
So those two cases of P2PSIP being used with HIP just work out of the
box, no interactions, and IMHO no layering questions.
In case (c), the peer protocol is taking the place of other rendezvous
and location services (DNS and a standalone HIP rendezvous application,
for example). I don't think there are any layering concerns here, it's
just a different application being used to accomplish the same service.
For case (d), essentially what you have is the peer protocol forming an
overlay on a partial graph of the HIP environment. The peer protocol
then uses its overlay as rendezvous server to add new edges between HIP
nodes (either to the overlay or for other applications). Is this a
layering violation? Well, without the peer protocol, HIP calls up to
DNS and a rendezvous server to form connections. With the peer
protocol, the overlay provides those two services. In both cases, an
application layer service is being used to provide help to form the
connection.
Now, is it a problem that messages may be sent over a HIP connection to
provide this service to HIP? Honestly, to me this looks an awful lot
like OSPF, which uses IP encapsulation when sending packets between
adjancencies.
Personally, I think from a purist's point of view HIP and P2P are both
types of overlay networks, which naturally encompass some amount of
layering violation if you take a pedantic view of the layers. I don't
find that any of the methods of using HIP with P2PSIP present
substantially new layering issues.
Bruce
_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip
