[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RRG] Long term clean-slate only for the RRG?



In einer eMail vom 04.07.2008 03:09:29 Westeuropäische Normalzeit schreibt brian.e.carpenter@gmail.com:
For an algorithm it doesn't matter 
> whether the network is tightly meshed or lesser meshed.It would only matter, if 
> there were no single mesh at all.

I'm sure that statement is true as a matter of graph theory. But we're
concerned here with scalability of practical distributed routing
algorithms, and meshiness has a big impact on aggregation. Since the
relatively small amount of meshiness seems to be a result of the
interaction of BGP4's design with the free-market economics of ISPs,
it isn't a parameter we can tune at will. (I think that's the basic
problem many people have with NIRA or *any* geographic scheme - it
seems to assume arbitrary meshing in a given geography in order to
scale well, and the Internet isn't like that.)

    Brian

Well, an architecture is more than a single aspect and "my" architecture is not based on DV like BGP but based on Dijkstra and even on better usage of Dijkstra. You may  utilize geographic data for the sake of DV as well, but this is not my thinking at all. More general: You may utilize  geographic data in many ways - in more as well in less obvious ways.
 
The mentioned little meshiness is due to the fact that the intra-domain ISP network is not adequately integrated into the inter-domain view (or better said: not at all).
 
Again, architecture is more than one aspect and research should start by questioning the old paradigms.
Is Distance-Vector sacrosanct or not ?
 
Another analogy :-)
Modern soccer is  "one-touch-only ball forwarding". What about "one-touch-only packet forwarding" ??!!
 
Heiner
 
 
Heiner