The IP Over NBMA (ION) Archive

[Albert Manfredi <manfredi@engr05.comsys.rockwell.com>]

I think that draft-armitage-ion-venus-00.txt pretty thoroughly covers the 
problems of cut-through multicast over unrestricted size ATM nets. The 
issues are analyzed at the microscopic level, convincingly so, but less 
microscopic investigation would have come up with the same conclusion.

In general, all multicast models that work aggregate at different levels. 
This is true for "broadcast" radio and TV as it is for IP mrouters. So if 
we have an ATM network that spans the globe, there is no reason to 
believe that reasonable multicast can possibly be supported without the 
same sort of aggregation.

One solution is to say that at a certain point, we can't do pt-mpt VCs 
where the source must be aware of and maintain a list of all active 
participants, and must therefore resort to mrouters. To me, this is 
avoiding the issue "But how would a global ATM do this?" I don't believe 
the answer is "The global ATM must use IP multicast solutions."
 
Or if it is, then we've again given up on making use of this different 
technology's potential benefits.
 
The answer has to lie in using ATM's own aggregation capabilities, in its 
own topology. Assume that ATM is flat, and that type of solution will 
never happen.

Not a job for the IETF? True. It's a job for the ATMF, and hopefully the 
IETF would make use of those solutions.

Bert
manfredi@engr05.comsys.rockwell.com