In einer eMail vom 30.07.2009 16:35:24 Westeuropäische Normalzeit schreibt [email protected]:
On 30 Jul 2009, at 09:54, Eliot Lear wrote: > On 7/30/09 3:24 PM, RJ Atkinson wrote: >> >> On 30 Jul 2009, at 02:40, Scott Brim wrote: >>> I just wonder if they all should be solved with the same mechanism. >> >> I suggested that we needed an architecture that supports the >> broad range of issues before the RG. Eliot sought an example, >> so I provided *an example*. >> >> Now you and Eliot have both gotten bogged down in the example, >> and discussion of architecture seems gone from the thread. > > I raised engineering constraints that we need to consider when > designing > a system, and by no means all of them. Determining the point at which > that crosses to architecture should occur directly after we've counted > the number of angels on a pin. No. You saw it yourself. As soon as the phrase "engineering constraints" appears, one has left architecture behind and crossed over into engineering. Mind, it is very important to get the engineering right. The Internet community keeps learning the hard way that it is also very important to sort out the architecture BEFORE crossing over into engineering. This is the kind of thing that jnc and some others of us have been saying for a few years now on this list, so it is neither a new thought nor uniquely mine. Yours, Ran IMHO, the problem with the RRG is that it is completely backward oriented. Today each handphone is able to take advantage of the geographical coordinates/GPS. But routers of even a future routing architecture still must not. Ran pointed out that scaling is not all. Indeed, I have also stressed many times the many advantages by viewing the internet topology. 1-2 years ago I asked the rtgwg-folks whether their first accomplishment was the end or the beginning. The answer I got: the beginning. But since then nothing occurred. Multipath forwarding can be substantially extended, not just ECMP and not just what the rtgwg did, but as well detours which even start out with crankback, entirely connection-less, - yes, intra-domain wise But in inter-domain routing there is not even a rear mirror. Hence neither a COMPLETE Multipath forwarding nor adequate TE (as to deal with traffic congestion) can be done. Will say: intra-domain routing could be improved substantially, but in the same sense inter-domain routing, too. I could repeat the entire list again and again what is prevented so far.... Since quite a while I realized that it will take some years till the LISP worship is over. So I keep being patient but allow me to speak up whenever this group gets stuck again with the same old problems. I remember the July meeting 2000 in London. It was said: well we have this problem, but no rush, we have a ten years time frame. Now we have July 2009. A second reply wrt to the "users without home agent" point. At first, it is only relevant that just one, i.e. the destination is without home agent. And yes, a reasonable service could be to search for this user, in the case where the source has only an idea inside which geo-patch the destination might currently be located. Heiner
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
