> From: Lan Wang <[email protected]> > we conclude that, to effectively solve the routing scalability problem, > we first need a clear understanding on how to introduce solutions to > the Internet which is a global scale deployed system.
Well, exactly. I had some thoughts on how to deploy new stuff in a "global scale deployed system" (nice phrase), which may be of some interest, here: http://www.ietf.org/mail-archive/web/rrg/current/msg04037.html ("what's _most_ important about any replacement piece of the internetwork architecture is not how nice a design it is, but how good its deployment path is", etc, etc.) Without trying to make a plug for LISP, I'd like to point out that it is, in essence, basically the same as the deployment plan for Nimrod - keep the hosts unmodified, and have routers add a new header which contains a locator - and for many of the same reasons, chief among them practicality. (Albeit it's a somewhat stripped-down plan - since the new locator namespace is syntactically identical to the endpoint identification namespace, unlike in Nimrod where they were different.) Although I must say, working on LISP has made me aware that the difficulty of that sort of deployment plan is even larger in practise than it was in theory... (as SCrocker so wisely observed :-). (Also, even though LISP in itself does not introduce any new routing mechanisms - although in separating location and identity it may allow us to attack some routing problems - I am still hopeful that if and when LISP is deployed that it will serve as a springboard for the deployment of a new locator namespace and a new routing architecture - and it has hooks built in to allow some of that.) The extreme importance of a viable deployment plan can be seen in the fact that IPv6 has so far not seen much uptake because it didn't have a good (along all axes, including economical) deployment plan. Noel _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
