Re: [arch-d] ETSI launches new group on Non-IP Networking addressing 5G new services

[Brian E Carpenter <[e-mail]>]

On 10-Apr-20 12:09, Toerless Eckert wrote:
> On Thu, Apr 09, 2020 at 03:07:14PM -0700, Dino Farinacci wrote:
>>> I was the one who asked the embarassing question about SRv6.
>>
>> And I was the won who said ???are we going to leave any data for the user???. Where I heard a loud Tony Li laugh from the audience.  ;-)
> 
> Btw. There are interesting hybrid optical/electrical research
> network architectures, where optical packets arriving are
> passed through an optical delay loop and the packet header
> spoofed into the electrical side to make packet
> processing decisions, which can then happen when the packet
> header starts to appear at the end of the optical delay loop.

That's roughly how computers worked in the early 1950s, with words
stored serially in mercury delay lines, where the bits travelled
at the speed of sound, so that micro-second latency vacuum tube logic
circuits could comfortably do their thing before the bits emerged
from the delay line.

Such computers were almost impossible to program and they died as
soon as faster memories were invented. The problem was that you had
to be ready exactly when the bits you wanted came around; this was
called optimum programming and the people who could do it were
the übergeeks of 1950.

As you said, "research architectures". Or limited domains, as
somebody else said.

> 
> Given how these optical networks will be many orders of
> magnitudes faster than the electrical processing
> capability, we may end up with networks that just scream
> to get larger packets to max out performance. 100 Kb minimum.
> And then we can waste all the header space we ever wanted.
> SIDs in URL encoding, x86 BIER replication function !!!!

Sure, but it won't help typical UDP streams much. We may not like
the historical traffic patterns, but we can't ignore them.

   Brian


> 
> Cheers
>     Toerless
> 
> _______________________________________________
> Architecture-discuss mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/architecture-discuss
> .
>