In einer eMail vom 26.07.2009 16:08:29 Westeuropäische Normalzeit schreibt [email protected]:
> From: [email protected] > By knowing the topology you can see all paths in forward direction wrt > a given destination as well as which network part would use you for > transit (the rear mirror). You have much less by assembling hundred > thousands of routes instead. True - but can we expect routers to have a _complete_ map of the network, or not? Of course not (this is not even required in a multi-area ospf network).Instead Hierarchy is the clue for coping with scalability. But there have been many many (failing) attempts so far : Nimrod, PNNI, ISLAY, Brian's, Rangi, CLNP, HAIR, LISP (2-staged), and now (N)-DIF. Also E.164.Non of these caters for a "sliding" hierarchy such that every node is (well, almost) in the middle of its nearest surrounding topology at lowest level, followed by more and more sparsed topologies at higher levels, the more they are remote. Instead they all are static, loaded with political and/or ISP-sensitive dependencies. Even leaving aside technical issues such as the amount of information, some sites may wish to 'hide' information about the internals of their network. Sure, and also while some other internals might be solicited to the outside. But these are well-known challenges. There are more in addition: How to enable such a sliding hierarchy. How to implement a hierarchy where strict links do not only show up inside the immediate neighborhood but interconnect, eventually, routers which are 20 000 km away from each other. >So then you have to work out how to make the routing work when entities do >not have complete maps... ... yes, and where there are also entities around which aren't aware of such a new technology at all. Noel Heiner
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
