In einer eMail vom 23.03.2010 20:35:15 Westeuropäische Normalzeit schreibt li...@cs.ucla.edu:
by Mark Handley, http://www.cs.ucl.ac.uk/staff/m.handley/papers/only-just-works.pdf In preparing our meeting for Friday, I found the views in this paper refreshing Lixia (no hat) Lixia, you find this view refreshing, but you nevertheless avoid questioning the existing routing paradigms: You don't see or admit the primitiveness of the distance vector algorithm compared with Dijkstra and the capability to determine a next hop based on knowing the topology. Yes, nothing has changed during the last 20 years of BGP-based inter-domain routing. I call it "progress by inzest" if TE is based on manipulating the PATH length by adding the same AS more than once. Nevertheless, I do not blame BGP, because I know that BGP could as well be improved as to become an inter-domain topology aware protocol. You can collect millions of routes and are nevertheless in a poor shape with DV, as it prevents knowing a huge number of routes due to its design. That cannot be improved -unless you dump DV! You cannot develop an adequate congestion handling mechanism if you don't know the topology! Obviously the paper you are referring to complies with the view that congestion handling is a network layer issue and not a transport layer issue. But the network can handle it only if it knows its topology! The RRG is supposed to be a research group, i.e. not a group which deals with the next moment's need. But I do not see that the dominant people even react on arguments that question the existing paradigms. You cannot win the future with DV - period. Certainly, Dijkstra is not sufficient, too, but Dijkstra can be improved! Substantially improved! However with the current attitude you won't accomplish any change for the better. You will go on with the most primitive TTL-mechanism ever. You will go on with the phantom fear wrt loops.You will go on with the orthogonality between intra- and inter-domain routing, and won't realize the advantages if that were overcome.You will keep off a whole generation of students from developing a network layer that deserves this name. Note, the current IP network layer wouldn't work without ITU-T's phone networks. IETF's victory over ITU-T was based on its advanced routing capabilities (imho). But today car navigation outperforms the internet's routing capabilities by far -without complaining about scalability problems, btw. Heiner
_______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg
