The IP over ATM Mailing List Archive by date

[fred@cisco.com (Fred Baker)]

At 5:40 PM 3/29/96, schulter@zk3.dec.com wrote:
>I'm not asking if IPs important to us; I'm asking if IP is important (enough)
>to the folks that developed ATM.  My hope is really that it is, but I
>have some nagging doubt.  When I see people state that "ATM must change"
>to better support some IP functionality, I can't help but wonder if such a
>desire is realistic.

Sooner or later, I'd like to see some realism/pragmatism set in. We've been
around this block often enough.

The fact is that IP has been moved to any number of different link layers,
and whenever a new link layer comes in (1822, X.25, bridged ethernets,
switched ethernets, frame relay, SMDS, and now ATM) there has been a
perception on the part of some that the users of the link layer were going
to somehow change to meet the link layer's self-centric view of the world.
Surprise, surprise, every single time we have this silly debate, and every
single time we find that the link layer network, whatever it's nifties may
be, is not going to replace the network layer any time soon. The
applications don't work right with a bazillion immediate neighbors,
multicast doesn't scale, you lose the addressing hierarchy, pick your
poison - each one has taken it, and each one, with the exception (to date)
of ATM, has come full circle and decided that in the end there needs to be
some coexistence. There is no realistic question of ATM obviating IP;
that's just not going to happen. There needs to be an adequate
understanding of the link layer service in the routing protocols that use
the service, and there needs to be some recognition on the part of the link
layer that the link layer is not the center of the world. The global
internet is not 1822, it is not X.25, it is not bridged Ethernets, it is
not Frame Relay, it is not SMDS, and it's not going to be ATM.

But IP, wherever you want to go, can get you there and will continue to do
so for the foreseeable future. That is not IP bigotry, that is a fact that
is demonstrable with a phone call to any service vendor you want to call.

Think X.25. World-wide, it has a definite place in the sun. What do people
do with it? Within France and a couple of other relatively small regions,
they run X.25-specific services; everywhere else, they put comparable
services on the network layer and do the same thing on a global scale.

Think Voice on Frame Relay. You can get a FRAD that runs voice across it,
but I'll bet you don't find that in the average home or in the average PC
catalog. What has wide market appeal in voice on RTP/UDP/IP, which can run
from your PC wherever it is to my Mac wherever I happen to be.

ATM-specific applications will have their place. But application developers
that can read history will put them on IP on ATM, and make sure that IP has
the smarts to use ATM well, which is ultimately going to mean getting
RSVP/ATM interactions right.

Call me obnoxious, call me pig-headed, call anything but late to dinner.
And if you want wide marketplace acceptance, smell the coffee.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
It has recently been discovered that research causes cancer in rats.