What are "participating transit intermediaries"?
To me, an intermediary is anything that handles network traffic and
which is not an endpoint, regardless of what layer of the stack it
operates on, or on whose behalf it is operating.
Do the charter authors intend that this group's purview include bridges,
routers, NATs, proxies, firewalls, gateways, etc? how about things
that operate on multiple layers ?
I think, and you might or might not agree with me, that OPES is a reaction
to evolution that's occuring in peer-to-peer applications. In a general
sense, there is a need to be able to embed "transit intermediaries" to
transform and operate on a message flow from a source peer to a destination
peer. Think of it not as peer-to-peer applications but peer-through-peer
networking that provide store-and-forward characteristics.
OPES, it seems to me, is trying to deal with a legacy client-server system
and transforming it into something that looks and tastes a bit like a
store-and-forward peer-through-peer networking system for X-over-HTTP or
interneworking between cache/content-delivery systems. The WG might uncover
interesting issues that would be useful for future peer-through-peer network
systems.
Architecturally, it seems to me that there are three components in a
"peer-through-peer" system forwarding layer, if location-independent naming
is used to identify remote peer resources:
a) topology construction to create "peer subnetworks" and interconnect peer
subnetworks
b) routing to propagate names and name space region reachability,
c) forwarding based on names from a source peer through transit peers to a
remote peer or peers
In such a system, interception and redirection are explicit ("I'm forwarding
this to a next-hop peer because they are on a path..."), and OPES recognizes
this. It's also desireable for the transit peers to be functionally
transparent, from the perspective that new types of peers can be added as
transit peers or intermediaries without changing the upper layer application
protocols. OPES assumes HTTP as a lingua franca that can be bent (as it has
been) to accomplish functional transparency.
I would tend to agree with you that that time might be better spent on
defining a peer-through-peer networking framework that defines a general
purpose forwarding layer for peer-through-peer systems. I do have some
thoughts in this regard that I could BOF. But there may be some utility on
OPES that uncovers the challenges for legacy systems.
-scooter