Cell Relay Archive

[Jonathan Turner <jst@cs.wustl.edu>]

Paul Koning wrote:
> 
> >But then IP often forces one to do
> >         really strange things (like NAT, for instance).
> 
> IPv4 does; IPv6 does not.
>
	IPv6 seems to be caught in the same never-never land as
	ATM. While a big improvement over v4, it contributes
	nothing with respect to QoS or multicast.
> > >
> >         The trouble with ATM in this role (as opposed to just
> >         directly connecting big IP routers directly over SONET)
> >         is that as long as ATM is subservient to IP, it adds
> >         no value to the network, only cost.
> 
> I wonder if you're being lead places by non-technical issues
> as expressed by adjectives like "subservient".
>
	I don't think so, but perhaps the term is too strong.
	Subordinate to IP is better, I suppose. And I think
	that term is justified. As long as IP defines the API
	and ATM is just a sublayer technology on which IP runs,
	ATM can offer little in my optinion.
> 
> As far as I can see, ATM definitely does add value to the
> network.  For one thing, it offers a large range of possible
> bandwidths because it's layered on top of the SONET hierarchy.
> (Admittedly the same goes for POS, but ATM has a significant
> head start.)
>
	Exactly. And POS is the competition that matters, not
	SONET. Unless ATM can play a bigger role than supporting
	IP in ISP backbones, it will become marginalized, and
	eventually squeezed out by POS (perhaps as soon as within
	5 years).

Jon