Hi Heiner,
If we are going to introduce a big change in the Internet Architecture (and Loc/ID split -is- a big change) to think that todays protocol will work with the same degree of efficiency is kind of utopia.
Once loc/ID is introduced, probably something like a future TCP-Vegas++ or TCP-New-New-New-Reno will fix the inefficiencies.
In other words, we should aim not at preserving top performances, but to not "degrade them too much".
Luigi Iannone Given these additional concerns about loc-id-split, what is wrong with my NIRA architecture, which would, as repeatedly said and proven, reduce the routing tables down to 2-3 K, eliminate the so-called update churn, and (!) speed up the IP forwarding by factor 20
I get the point of how NIRA would reduce the routing tables, but how NIRA can speed up IP forwarding on routers' line cards by a factor of 20?
(while providing a new and better basis for QoS/policy routing, traffic balancing,network maintenance,and overcoming the orthogonality between intra- and inter-domain routing) ? Heiner
|