[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ih] internet-future mailing list ?

Great insights. Inline.

On Mon, Dec 16, 2019 at 01:15:24PM -0800, Karl Auerbach wrote:
> It can be kinda fun to try to look into the crystal ball.
> One thought that struck me a few years back was a recognition that "The
> Internet" (with or without capitalization) is rather like the elephant in
> the fable about the blind men: it is perceived as many different things.

You do know it actually is not an elephant, right ?

>From a presentation i watched last friday:

> For those of us here "The Internet" may be conceived as a system that
> carries IP packets from hither to yon where that hither and yon are
> identified by globally unique IP addresses.

I think there are a lot who could think beyond that past defined box.

> From that application-centric point of view things like "end to end
> principle" become merely a disposable detail of inner plumbing.? Does it
> really matter to Twitter users whether the underlying machinery is elegant
> and free of media transitions and proxies?

Just because billions of people drive cars without understnding how they
work doesn't mean you do not need people to think about the future of
how cars could/should work. Even beyond electric engines.

But i get the point that user/application interest would completely
outweigh any lower-layer discussions. I would consider any such
discussion a success and worst case one might need to carve out more
mailing lists to keep specific groups interests focussed.

Right now i would not be worried about that.

> And from another perspective I've seeing a lot of movement, often done under
> the banner of "optimization", back towards circuit switching notions - or
> rather, hybrids in which packet routing is ever more forcefully constrained
> into fixed paths (especially for data flows for conversational audio or
> interactive video that have severe latency and jitter constraints.)

Yes. Or from distributed to centralized or at best little decentralized
due to hierarchical "SDN" style orchestration.

> And might one consider the 5G movement (even without millimeter wave
> technology) as a new ISO/OSI (but better designed to co-exist with existing
> IPv4/6 infrastructures.)

The non-radio side of 5G is just a distributed overlay application. It
does today AFAIK not cover the actual management of physical underlay
resources. Thats one of the big contention points going for who will
design that. The way i see it, 5G/6G underlays that do have good
loss/latency/throughput management could easier be built with IEEE 
technologies than IETF technologies. Unless we would start to think how
to move beyond our current understanding of how to build the Internet.

> A few years back I wrote up one view of where the net could be going.? It
> was somewhat pessimistic.? However the intervening years have not adduced
> much evidence to the contrary.
> https://www.cavebear.com/cavebear-blog/internet_quo_vadis/

Yes, very nice overview. I probably agree with a lot ow what you wrote
there, but haven't thought hard about all of them.

IMHO, the Internet will evolve into Fronthaul connecting devices&users
to edge-DC and backhaul interconnecting edge-DC. That evolution takes
out most or all problematic transit issues from the frontaul, therefore
enabling a lot more interesting network services than just best-effot
much more general than those are today. Will it be possible/desirable to
build new business around it. Thats the better question.

> One of the more interesting aspects of my own delving into Internet history
> has been that there were many roads not taken. Some of those roads could be
> re-explored.? (My own favorite candidate for that would be to revisit what
> the ISO/OSI people did so badly that few comprehended its value: a
> persistent session layer above transport.? Had we had that we would not have
> had to explore inelegant things like mobile IP or HTTP/S cookies.)


Internet-history mailing list
Internet-history at elists.isoc.org