Re: Other use cases for header insertion (was Re: Header Insertion and TI-FA)

I would love to hear i am wrong, but i am sure a new IP version number would introduce
more need for HW-upgrades. While i appreciate architectural cleanlyness, i really would love if
we would reserve the need for such fundamental HW changes to more fundamental improvements
to our IPv6 networkinging layer.

For example: IPv6++ could support different adress spaces of different address sizes, one
of which could be Internet/IPv6, another one Internet/IPv4 and many more for various address
sizes for the probably millions of private IP networks on the planet, including the "transport"
layer "network" of most SP networks with likely << 128 bit address sizes or even smaller IoT networks.
Interop between different address spaces/sizes is designed upfront instead of 28+ IPv4/IPv6 transition
solutions.  Addresses would not have to be (ab)used for every path processing function because the
header provides ways to define better alternatives and those are correctly MUST requirements in the
spec, and steering would not be constrained to the concepts of a 40 year old source-routing design.
Oh, and QoS would get more than 8 bits so we would not have to fight over semantic of 1 ECN bit.

To me, that would be evolution goals for which it would be worth to consider mayor HW upgrades and
assigning a new IP version codepoint. Would be nice if more IPv6 supporters would think about
these type of now well feasuble and understood long-term innovation options instead of only
trying to stick to the status quo in the hope of increasing adoption.

Cheers
    Toerless

On Tue, May 12, 2020 at 02:45:27PM +0000, Ron Bonica wrote:
> Robert,
> 
> I realize that you are joking, but it might actually be a good idea for SRv6 to move onto another protocol number. This would solve the following problems:
> 
> 
>   *   SRv6 could do whatever it wanted, without regard for RFC 4291,  RFC 8200, or anything else that 6man produces
>   *   6man could evolve IPv6 as it sees fit, without having to consider the sensitivities of SRv6
> 
> It would only mean changing the first four bits of every SRv6 packet. That would probably take much less time than converging the SRv6 and IPv6 architectures.
> 
>                                                          Ron
> 
> 
> 
> 
> Juniper Business Use Only
> From: Robert Raszuk <robert@raszuk.net>
> Sent: Monday, May 11, 2020 4:43 PM
> To: Brian E Carpenter <brian.e.carpenter@gmail.com>
> Cc: Tom Herbert <tom@herbertland.com>; Ron Bonica <rbonica@juniper.net>; 6man <6man@ietf.org>
> Subject: Re: Other use cases for header insertion (was Re: Header Insertion and TI-FA)
> 
> [External Email. Be cautious of content]
> 
> 
> I'm not sure this is feasible without changing the IP version number.
> 
> How about IPv6+ ?
> 
> Thx,
> R.
> 

> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> ipv6@ietf.org
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------

-- 
---
tte@cs.fau.de