[v6ops] Re: [E] New Version Notification for draft-mishra-v6ops-variable-iids-problem-statement-01.txt

[Lorenzo Colitti <[e-mail]>]

Brian: I think you are missing the point that the *flexibility* itself is
the problem here. "Only /64, ever", would be fine, "only /80, ever" would
also be fine (well, assuming you believe that that's enough space to create
random IIDs). The problem is that:

   1. Flexibility leaves it up to the operator.
   2. Some operators will respond to business incentives to provide the
   lowest-possible amount of space. (We know this because some operators are
   already doing so.)
   3. Software is forced to support lowest-common-denominator, and there
   are strong incentives *not* to support anything better than
   lowest-common-denominator, because doing so requires more testing.

So, unfortunately yes: while it's not a bad idea in theory, it *is* a
terrible idea in practice.

In case you're thinking of responding along the lines of, "oh operators
will do what they want anyway, regardless of what the IETF says": that's
not true. If it were true, we wouldn't be having this conversation every
few years. Instead, some operators would just have gone and done this -
after all, they've had plenty of time to do so. By and large that hasn't
happened: residential users still get /64 or shorter, and mobile devices
still get /64s (which, I might add, has allowed lots of permissionless
innovation like 464xlat and IPv6 tethering without NAT). The reason it
hasn't happened is that doing so doesn't work in enough cases that they
can't do it.

On Fri, Sep 20, 2024 at 5:34 AM Brian E Carpenter <
[e-mail]> wrote:

> So, my question is that if people have concerns about over-enthusiastic
> use of IPv6's "infinite" address space over the very long term, why is
> making the /64 boundary more flexible by a small architectural adjustment
> (already consistent with RFC 4861 and 4862 and BCP 198) such a terrible
> idea?
>
> Regards
>     Brian Carpenter
>
> On 20-Sep-24 05:21, Jeremy Duncan wrote:
> > " If you are also required to give a /64 to every device that wants to
> use DHCP-PD, and your ISP only gives you a /56, then that would limit you to
> > <256 such devices in your network."
> > **that's wrong. It's 256 /64 subnets
> > **but if doing DHCPv6-PD it's 256 devices that are doing things like
> containers, etc - whole other standards on this.
> > **there's no ISP that will do DHCPv6PD that will force /64 per device -
> what evidence do you have here?
> >
> > -Jeremy
> >
> > -----Original Message-----
> > From: Brian Candler <[e-mail]>
> > Sent: Thursday, September 19, 2024 1:11 PM
> > To: Jeremy Duncan <[e-mail]>; Gert Doering <
> [e-mail]>; Mark Smith <[e-mail]>
> > Cc: Lorenzo Colitti <lorenzo=[e-mail]>; IPv6
> Operations <[e-mail]>
> > Subject: Re: [v6ops] Re: [E] New Version Notification for
> draft-mishra-v6ops-variable-iids-problem-statement-01.txt
> >
> > [EXTERNAL] Verify links and attachments with sender.
> >
> > On 19/09/2024 17:11, Jeremy Duncan wrote:
> >> That’s a disingenuous argument. It’s like saying we need to care for
> >> the amount of sunlight we receive because it’s going to run out… in
> >> like 2-3 billion years…
> >
> > It's being squeezed from all directions.
> >
> > A /64 prefix is needed for each layer 2 domain (switched network). Those
> only scale reliably up to a few hundred devices; more than that and you
> need a router with multiple prefixes. Hence the need for at least a /56 for
> end users.
> >
> > If you are also required to give a /64 to every device that wants to use
> DHCP-PD, and your ISP only gives you a /56, then that would limit you to
> > <256 such devices in your network.
> >
> > At the other side, larger ISPs are being allocated /16's, of which there
> are only 8192 in total available, giving a land-grab and a clear path to
> burnout before much of the world is even using IPv6 in anger.
> >
> > All this because someone thought it would be a great idea to use MAC
> addresses to assign host addresses autonomously, without realising at the
> time what the privacy implications were. If they had, and we kept DHCPv6,
> we could give each end user a /96, and they'd still have more usable
> addresses than the whole IPv4 Internet. But yes, that ship has sailed.
> >
> > _______________________________________________
> > v6ops mailing list -- [e-mail]
> > To unsubscribe send an email to [e-mail]
>