Re: I-D Action: draft-bourbaki-6man-classless-ipv6-06.txt

[Lorenzo Colitti <[e-mail]>]

On Tue, Apr 19, 2022 at 7:26 AM Brian E Carpenter <
[e-mail]> wrote:

> > "Lets make IPv6 easier for people learn and adopt by making its
> addressing as complex as IPv4's", said no one ever.
> >
> > IPv6's addressing should be as simple as will do the job.
>
> But it doesn't only have one job. draft-bourbaki- is about its job as a
> locator for routing purposes, and all it really says is what BCP198
> (RFC7608) already says: the routing prefix is any length up to 128 bits,
> and longest-match routing always applies.
>

That's not what it's about - at all.  Even the abstract makes it clear that
it's about addressing, not routing:

=====
   The last
   remnant of IPv6 classful addressing is a rigid network interface
   identifier boundary at /64.  This document removes the fixed position
   of that boundary for interface addressing.
=====

and later the draft it says it's *also* about routing:

=====
   This document also clarifies that IPv6 routing subnets may be of any
   length up to 128.
=====

I continue to object to this document on the grounds that it makes it
impossible to achieve feature parity with IPv4 without abandoning
end-to-end connectivity. In IPv4, any edge network participant can
endlessly extend the network by using NAT. IPv6 allows this too, but in a
different way: because the address assignment unit is fixed at /64, and /64
always has enough room for more addresses, any network participant can
extend the network by ensuring individual /128s are routed or bridged
within their network. If we remove the fixed boundary, then that is no
longer possible without sacrificing end-to-end connectivity which is a
major improvement that IPv6 can provide over IPv4. I'm sure many others
feel the same way, which is presumably why the draft did not progress the
last time.

If the authors really intend this document to be about routing only, they
should ensure the text states that plainly. I'm sure it would be less
controversial and more likely to gain consensus.