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

[Toerless Eckert <[e-mail]>]

Not sure what exactly you're disageeing with. "mayor" ?

I guess everything is relative. Creating something not backward
compatible counts as "mayor" to me, even if its just 96 more
address bits. 

21++ 32->128 bit migration mechanisms is IMHO a lot of good experience.

All very painfull of course. stackevo summarized a bit of it, but
don't think we've really tried to conclude on how to avoid / minimize
these issues if we every would have a chance to redo. I guess most
high level marketing always says variable length addresses for
next-gen, but there is certainly a lot more architectural/mechanism
to be worked out ...

Cheers
    Toerless

On Fri, Apr 10, 2020 at 09:08:00AM +1200, Brian E Carpenter wrote:
> > Yes, and 30 years ago we made a mayor new step (IPNG -> IPv6)
> > and learned a lot through it.
> 
> Actually, I disagree. IPv6 is a very conservative design that had
> exactly no effect on fundamentals. That was intentional. Remember
> TUBA (TCP and UDP over Bigger Addresses)? That's what we did,
> with the tiny detail of using homebrew IPv6 instead of ISO8473.

> What we learned is that even a minor step like bigger addresses is
> expensive *because of the even more tiny details* like the socket
> API, router CLI commands, and the address configuration mechanism.
> 
> The major lower-layer atchitecture change in the last 30 years
> was completely unplanned: middleboxes.
> 
>     Brian