In einer eMail vom 17.04.2010 03:57:17 Westeuropäische Sommerzeit schreibt
scott(_dot_)brim(_at_)gmail(_dot_)com:
Unfortunately there isn't much new under the sun. This appears to be a
combination of geographic addressing (various sources, e.g. Steve
Deering and Tony Hain) and RFC1955. Geographic addressing has
deployability issues -- search for archives of those arguments. The
problem with mapping based on ASs is that they are the wrong granularity
for external routing. They are either too coarse (different prefixes
within an AS need different treatment) or too fine.
Scott,
there is a lot new under the sun - from the routing technologyical point of
view. In 2003 I really made a significant discovery that extremely
propelled all my routing skills. Either me or someone else, who however would
progress based on my work, will be able, sooner or later, to prove that P=NP.
The practical( i.e. less academical) benefits could be harvested even now,
but obviously no one from the RRG is interested in routing technology that
is more advanced than what you can learn at the universities.
Instead people rather rediscover Steve Deering's metropolitain routing.
But just using the geographical coordinates in that way won't do! This is as
poor as is ILNP which in indeed just a remake of PNNI and - remember you
created the acronym yourself- of IARP (Inter-domain ATM-network Routing
Protocol). In the meantime others have shown quite some negative aspects of
such kind of hierarchical routing, more precisely, of such hierarchical
address-summarization based routing: Stretch 17 !
And I can also add: Istanbul effect (exact city map for the European part
plus a cut-out for the Eastern part from a road map about whole Asia) ! And
third: No sensitiveness for a future, entirely mobile network!
What if A L L users become mobile ? including
the home agents and care-of-address servers ?!!! ILNP won't be of any help
at all.
And: ILNP emphasizes that multi-homing is just a special kind of multipath.
Great. But at the same time it prevents 100 % multipath, where there is no
panicing about loops anymore, not even in case of required crankback.
Heiner
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf