In einer eMail vom 14.07.2009 19:47:16 Westeuropäische Normalzeit schreibt [email protected]:
On Jul 12, 2009, [email protected] wrote: > Let's put it differently: Incremental deployability requires still BGP > to overcome the main function of BGP which is the non-scalable way to > disseminate all these hundred thousands of prefixes. The methods are > not > new: Remember BGP-based dissemination of MPLS-VPN sites (just lacking > the goal to build ONE SINGLE "internet-VPN") : here the information > are > labels, i.e. not IP addresses. Also remember LISP's default mapper. > Issueing a prefix of length 0 might attract traffic destined for a > node > which doesn't propagate its traditional reachability prefixes anymore > but just new routing info for a new architecture, analogously to the > MPLS-labels. Heiner - I agree with the argument that IP addresses could theoretically be replaced in their role of locators. Although I would add that the feasibility of this heavily depends on the deployment model: Absolutely. The model must a) work, b) be incrementally deployable and c) solve all the problems that are due to the state-of-art though bad architecture: abolish the scalability problem (even if the internet were a millions times bigger),overcome the IPv4 address depletion issue, cater for multihoming, make Moore's law become applicable and finally enable much better TE by also providing the view to the rear ( who else would use THIS router for transit to get to some destination down in the forward direction.) I wouldn't emphasize all these objectives if my model didn't enable them :-) LISP proxies are not a good example because they require IP addresses as locators to route packets towards a proxy. On the other hand, APT's deployment model -- which calls for initial local deployment and subsequent expansion -- could facilitate the introduction of new locators. Of course, those new locators would have to be suitable for use across administratively discontinuous providers, like IP addresses. LISP is an attempt to rescue the existing model and the existing paradigms.(called jack-up model) While most people praise the existing paradigms and e.g. think that caching is a nice enhancement thereof I rather believe that caching is a built-in repair system of an old/bad architecture from the start. Heiner - Christian
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
