Noel and others suggested we might be possible to reach consensus on: > 'The RRG did reach a rough consensus that it is both desirable to > separate location and identity, and also technically feasible to do > so.' > >> As long as you don't try to define identity. > > Which is part of why I stuck with the generic concept terms 'location' and > 'identity', and didn't use any specific examplars of either. > > But seriously, if anyone disagrees with that concept, let's hear it, > otherwise I would like to ask Tony to include it as a 'group endorsed > high-level recommendation'.
But what does "separate location and identity" mean? If we can't agree on what it means and which architectures implement it and which don't, there doesn't seem to be any point in agreeing on it as a concept. HIP, GSE, ILNP, RANGI, GLI-Split and Name Based Sockets all implement "Locator / Identifier Separation". Hosts have both an Identifier (or perhaps multiple Identifiers) and one or more Locators. This requires stack and perhaps application changes from the current stack and app arrangement where the IP address of the host functions as both its "Locator" and its "Identifier". If that is what it means, then I do NOT believe it is the way forward to solve the routing scaling problem. My main objections are that it places too much extra work, extra packets etc. on individual hosts - and that it typically introduces extra delays in establishing communications. I think this is a bad deal in return for keeping the network simpler. I think the network should gain a new layer of complexity along the lines of a Core-Edge Elimination architecture (Ivip, IRON-RANGER or LISP), and that hosts should not gain any extra complexity or responsibilities. Comprehensive arguments and references are here: http://www.ietf.org/mail-archive/web/rrg/current/msg06219.html I am concerned that the phrase "separate location and identity" is believed by some people to apply to LISP - and LISP, Ivip and IRON-RANGER share many major architectural features. LISP, Ivip and IRON-RANGER does not involve hosts in any changes, so they do not involve hosts at all in "Locator / Identifier Separation". Some argue that within LISP, there are Locators and Identifiers - but I think this is contorting the meaning of these terms. See the thread: LISP does not implement Locator / Identity Separation http://www.ietf.org/mail-archive/web/rrg/current/msg06190.html So if we can all agree that "separating locator from identity" refers to doing so at the hosts, then this phrase does not cover Ivip, LISP or IRON-RANGER. Then I would NOT be an RRG participant which was part of any rough consensus agreeing to: > 'The RRG did reach a rough consensus that it is both desirable to > separate location and identity, and also technically feasible to do > so.' Nor would anyone who favors LISP, Ivip or IRON-RANGER. Then, I expect, we would not have consensus support for this concept. If there is consensus that LISP is also an instance of "separating locator and identity", then I won't cast a vote against this concept having consensus support. However, I think it is a meaningless concept if it doesn't clearly define that this separation occurs at, and directly effects, hosts (GSE, ILNP etc.) or whether it also applies to architectures which do not alter host responsibilities, stacks, apps, communications etc (LISP, Ivip etc.) but which are argued by some to include "separating location from identity" inside the new things they add to the network. - Robin _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg
