On 2010-10-22, Dmitri Krioukov wrote:
the lack of consensus is quite expected.
Of course. A research group charts/hashes out possibilities and then in the end tries to focus real-life protocol development along the seemingly most productive avenues based on that. IRTF exists separately from IETF because somebody does need to intelligently speculate at the grand scale and using pure results from the academia as well. That sort of exercise isn't concensus bound, unlike the engineering side done within the IETF. Hence the division.
it occurred to me in some off-line discussions here, that folks are still looking for scalable routing which would work for any possible network topology.
AFAICS, we're now working two separate topologies: the logical/identifier one, and the more well-known topological one. In isolation from one another, and needing some means of efficient and transparent mediation/mapping between them.
[...] scalable routing for arbitrary networks cannot exist. in fact, it's intuitively easy to see, but it's also a proven fact, like a theorem.
It is. But even the charter of the RG has already moved past that dilemma: now we're considering architectures where there isn't just a single common routing architecture, but at least two (end point identifier vs. topological location) which are dynamically mapped via some way. We're not trying to do what you think we should do. We're doing something else entirely, using some sort of mapping between two namespaces, on the fly. The real debate is now about where to do the mapping, how to map it into the existing IP infrastructure, how best to do the mapping in a router and the interoperability and especially about how to distribute the map itself efficiently enough to make this whole thing transparent.
the only option is to look for solutions which would work only for *specific* networks, e.g., the global internet
That is not true. If you claim that, you will have to prove it beginning with first principles. And I think it's already quite clear you cannot do it using the basic axioms of the Internet, so to speak. ;)
But you are still quite right in that a) if we have a reason to believe the global Internet has some invariants, with strong theoretical reasons why we should not expect them to change, then b) we absolutely should find ways of exploiting them for routing efficiency, and whatnot.
If you can find such invariants in topology and can also tell us why they would hold over time, then you should definitely suggest a rechartering of the research group. Right now the RG is busy and also formally tasked with much simpler things. Why don't you help with the simpler thing first, while submitting a recharter request in parallel, so that we have something new to study as soon as possible after the current exercise?
such solutions may be efficient and scalable if they find a way to intelligently utilize network peculiarities.
Agreed. At least MANET's and the like have done this sort of thing for awile, and we even have IETF WG's working on this stuff. But that's no good reason to detract our attention from the current architectural problem, which is still unsolved and not ready to be fed to IETF's protocol designers.
Just as a hint, I've followed IETF and then IRTF work from afar since my early teen years. Many other standardization organizations as well, like W3C, IEEE, ISO and the like. What always impressed me in the IETF work (and IRTF by extension) was how goal-oriented, factual, free, and to the point it was. You too, just let it fly to its conclusion, and then suggest something new. In the meanwhile, contribute actively so that we might have something new to research/study over the next cycle. Help put the current topic to rest.
-- Sampo Syreeni, aka decoy - [email protected], http://decoy.iki.fi/front +358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
