While Noel's critique completely aimed into the blue, Lixia's compromise is no help either. It makes no sense at all to use geographical coordinates together with distance vector algorithm: The fastest way to get from Wiesbaden,Germany, to New York, USA, is to drive EAST !!! to Frankfurt airport as to fly from there WESTBOUND. However it makes a lot of sense, to attach geographical coordinates to the nodes of the internet (more precisely to the xTRs of the internet). While storing&updating the entire topology wouldn't scale, 5 different zooms (see Google maps) would. All it takes is a BGP extension whereby a BGP-router advertises a TARA-link which comprises : the zoom-level, the 2 adjacent nodes(IPv4-address + TARA-locator), the link weight (number of hops) and an information whether it is a TARA-link or (importent during the incremental deployment phase) a GRE-Tunnel. This is very little compared to LISP and others. But again, such a compromise (geo.coordinates combined with DV) will truly compromise i.e. discredit the whole idea. Heiner In einer eMail vom 02.02.2010 08:12:11 Westeuropäische Normalzeit schreibt [email protected]:
On Jan 14, 2010, at 7:54 AM, Noel Chiappa wrote: > Ah, no. > > Everyone who keeps going on about embedding geographic information > into the > names used by the path-selection is missing something really critical: > > ***Two computers which are _across the street from each other_, in > geographic > terms, may be (and often, are) _many hops apart_, in network terms - > because > they are connected to different ISPs whose geographically nearest > point of > connection is a long way away (e.g. in another city).*** > > Geographic information about two computers tells you _nothing_ about > how close > they are to each other, in terms of the path through the network > between them. > That is why the names used in path selection have to be based on, > and embody, > only the _actual network connectivity_. > > Now, can we stop being hearing this ridiculous nonsense about > embedding > geographic information in the names used by path-selection? > > Noel sorry for this belated reply: Noel, looks like you missed part of the picture being discussed here. I fully agree with your comment if you talk about solely geo based addressing. But if you look at the msg subject: it is about coding the ISP/AS# as *first* part of the address, and having geo info only *after* that. In this case the above comment does not seem applicable. Lixia _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
