Hi Dan, Andrew, Preethi,
This is in response to your tsvarea presentation today and your draft-
wing-http-new-tech-00.
As I commented after your talk, I think the whole negotiation topic
that you brought up is incredibly important. I think the specific
approach you're proposing is very sensible and perhaps really the only
realistic option in the short term, and certainly worth pursuing
further, likely in the application area.
In the long term as alternatives explode, I think we'll need something
more general and scalable, and that can deal with both transport-level
and application-level negotiation. Jana and I actually just recently
wrote a draft paper on this topic - our version is pie-in-the-sky and
researchy, but I hope it might be interesting as a potential longer-
term approach to this problem. (And your short-term approach will
needed to get to such a long-term approach. :) ) Here is the title,
abstract, and URL:
"An Efficient Cross-Layer Negotiation Protocol"
http://bford.info/pub/net/nego.pdf
Internet evolution often depends on either inserting new protocol
layers or upgrading existing layers to new protocols, but both of
these evolutionary paths are obstructed by the difficulty and
inefficiency of determining which protocols a pair of hosts mutually
support and prefer. We propose a novel cross-layer Negotiation
Protocol that sets up a complete stack of connection-oriented
protocols at once, concurrently performing handshaking for multiple
layers and choosing among alternative protocols for each layer in as
few round trips as possible, often just one. The initiator proposes a
protocol graph explicitly encoding possible configurations along with
protocol-specific handshake data; the peers then prune, refine, and
atomically commit to a final configuration, exchanging messages over a
specialized transport that can operate in-line with the negotiated
protocol stack. Although a practical Negotiation Protocol presents
many challenges, our initial exploration suggests that these
challenges are solvable, and we believe addressing them is a necessary
step toward a more evolvable Internet.
We might try to talk about these negotiation ideas tomorrow at the TAE
Bar-BOF; see my previous message on tae and tsv-area...
Thanks,
Bryan
