On 19/03/2019 13:08, Ted Lemon wrote:
On Mar 19, 2019, at 8:05 AM, Stewart Bryant <stewart.bryant@xxxxxxxxx
<mailto:stewart.bryant@xxxxxxxxx>> wrote:
Rather than consider this aspect of our work, we should put more focus
on putting in place the telecommunications technology we need to have
in place in the 10 to 20 year time-frame. By then we will need
bandwidths in the 1TB class to the users, power consumptions per bit
that is minute compared to the current level, and latencies where the
laws of physics are the only significant factor. It is unclear whether
the current Internet architecture will satisfy those needs without
change, and in my view that is the area that the IAB should attend to,
rather than to work described in this draft.
This is certainly an exciting picture of the future that you are
painting at layer 2 and layer 3, but I notice that you didn’t respond to
Ole’s point, which I think is a great example of the problem. Like it
or not, the participants in the IETF have expertise that is relevant to
things that you call “political,” and to suggest that we should not
factor that into our decisions is unrealistic. We are going to factor
these considerations into our decisions whether we admit it or not.
But where IETF discourse goes badly wrong is when we have discussions
where we haven’t agreed first on what we are talking about. Then you
see these hundred- or thousand-message streams of people talking past
each other and never reaching completion, because nobody feels heard,
because we didn’t take the time to figure out what questions we were
trying to answer.
If you think about it, what you have proposed is actually just as
political as anything else. You believe that users should have
maximally fast, minimally latent connectivity. I think that’s a good
goal, but it’s only a meaningful goal if the users want it. Otherwise
it’s just some shiny tech you’re working on. And let’s be clear about
this: it’s not at all clear that users want that. What I mean is that
users aren’t making purchasing decisions on the basis of whether they
maximize throughput and minimize latency. If they were, we wouldn’t
still be seeing bufferbloat everywhere.
Understanding what users need and trying to address their needs is in a
sense the core function of the IETF. It’s not a side issue that we can
ignore in the pursuit of technical excellence.
Just to be clear, those are not numbers I plucked out of the air.
1TB/s is needed to do holographic teleconferencing. I would argue that
being able to have a fully immersive experience including shaking the
hand of your business contact, or hugging a parent or child without
burning carbon must be a good thing that many people will want to do.
The haptic component of the experience requires surprisingly low latency.
Then there are many other low latency demands such as people packets,
i.e. driverless cars.
Obviously you can do no better than the physics, but the latency and b/w
applies all the way up the stack.
Of course we have to do all of the above without switching the global
energy demand from transport and heating/cooling to networking and
computing.
Then it is clear that the sensor needs are only going in one direction
when it comes to quantity, energy, material, and silicon area in a post
Moore's Law world.
The above is the clear direction we need to be pushing, and yet it is
by no means the IETF priority.
If you want more information look at the work that 3GPP and ITU are
doing on 5G and on Future networks respectively.
There seems to be a good case that many things need to change in the
Internet, and it is really up to us whether we want to be part of the
solution or an extinct part of the problem.
Meanwhile, I really think we should focus on technology and let
governments worry about the highly complex politics of the Internet.
- Stewart
