Waste will kill ipv6 too
morrowc.lists at gmail.com
Thu Dec 21 16:48:18 CST 2017
On Thu, Dec 21, 2017 at 11:20 AM, Lee Howard <lee at asgard.org> wrote:
> From: <christopher.morrow at gmail.com> on behalf of Christopher Morrow <
> morrowc.lists at gmail.com>
> Date: Wednesday, December 20, 2017 at 6:07 PM
> To: Lee Howard <lee at asgard.org>
> Cc: Mike <mike-nanog at tiedyenetworks.com>, nanog list <nanog at nanog.org>
> Subject: Re: Waste will kill ipv6 too
> On Wed, Dec 20, 2017 at 2:16 PM, Lee Howard <lee at asgard.org> wrote:
>> I’ve tried several times to come up with a scenario that leads to
>> depletion in less than 200 years, and I haven’t managed it. Can you do it?
> during some ARIN discussions that revolved around Transition Technologies
> and allocations to large ISPs, there were more than a few folk batting
> around the idea that they may need to allocate a /24 or a /20 even to a
> single provider.
> I believe DT has a /19 assigned to them currently? how many /19's are
> there in the v6 space? (524288-ish)
> That's only ~100x the current number of active ASN in the field. It's
> unclear (to me) how many of those could/would justify a /19 equivalent, and
> how fast the ASN field is growing over time.
> DT is one of the largest ISPs in the world, isn’t it?
it's large, but really it's going to hit the same number of homes (about)
as att/verizon/comcast/embarq ... and I'm sure ntt, 'russian cableco', the
5 china-cablecos etc. Right?
germany is ~83m people... 100x that is about 1.2x world population, so ...
it seems conceivable that there are ~100 isps (one per country) about the
same size, right?
> Can you devise a scenario in which there are 524,288 ISPs the size of DT?
I think I did something in my reply which I should not have done, I
conflated the DT issue and the transition technology discussion...splitting
1) For DT, my understanding is that their allocation is this size due to
part of their deployment plan/technology.
(multiple /48's per site, one per particular technology in use - video,
voice, intertubes, on-demand-video, something...7/site I believe was their
2) For the transition technology discussion I believe it centered around
attempting to get a /48 to each 'site' (home/customer) and doing ds-lite as
the transition technology in use.
(map the customer to not a /128 in the ds-lite, but a /48)
> Or one where every currently active ASN, times 100X, needs/justifies a /19?
> 200 years seems optomistic, 20 years seems easy to imagine surpassing
> though. What's the sweet spot?
> 200 years seems pessimistic to me. Every scenario I run uses ridiculously
> profligate assumptions, and usually multiplies those by a few orders of
> magnitude. Even extrapolating from your math above, I don’t get less than
ok. I think a bunch of the analysis so far in this thread has basically
assumed dense packing at teh ISP and RIR level.. which really won't happen,
in practice anyway. I was simply stating that if we follow some of the
examples today it's no where near as certain (I think) that '200' is ok to
A larger point is: "so what?"
we've run a number conversion / renumbering once... we can do it again,
better the second time, right? :) Maybe this next time we'll even plan
based on lessons learned in the v4 -> v6 slog?
More information about the NANOG