The IP Over NBMA (ION) Archive

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Steve;

> The point I was making is this:  Ohta has argued that 1) multicast layer 3
> routing does not scale over large multicast layer 2 NBMA topologies.  From
> this point he concludes that 2) NHRP, a unicast layer 2 shortcut resolution
> protocol, is not useful.

So, I wrote recently to ION ML,

	Sure. It's mostly fatal. Moreover, it is not a bug. So, bug fixing
	is impossible. See draft-ohta*mcast* for details.

	NHRP can still survive in unicast X.25 environment.

But, are there real-world users? Isn't the Internet today is more
widely available than X.25 even in developping countries?

I don't think people accept the idea of unicast-only ATM.
Or, do you disagree?

> routers).  My claim is that point 1 does not preclude this sort of topology,
> and, hence, the scalability of layer 3 multicast routing has no bearing on
> the suitability of large layer 2 NBMA attached networks

That IP can't make use of a large layer 2 NBMA does not directly
mean non-IP users are forbidden to have a large layer 2 NBMA.

But, it has nothing to do with NHRP nor IETF.

> His argument only bears on the number
> of receivers served by a single multicast router, which is a LIS scalability
> issue, not a "large cloud" one.

If I understand you correctly, you are technically wrong here.

Your assupmtion, seemingly, is to:

	Have multicast router in each LIS for intra-LIS multicast.

But, the resulting multicast tree is badly inefficient.

If you somehow think inefficiency is not a problem, unicast
inefficiency is not a problem either. So, you don't need NHRP.

Finally, how can you enjoy end-to-end cell-by-cell relaying over
multicast routers with such configuration?

							Masataka Ohta