Waste will kill ipv6 too

Jimmy Hess mysidia at gmail.com
Thu Dec 21 14:50:06 CST 2017


On Wed, Dec 20, 2017 at 3:57 PM, Mark Andrews <marka at isc.org> wrote:
[SNIP]

25B  estimate for earth's carrying capacity for humans is likely on the
high side,
but sure: IPv6  should suffice  until  we have a few planets' worth of
humans,
and require an  interstellar   IP network  with end-to-end  comms between
every
remote device in our galaxy cluster  ---   and may have to fallback to
planetary NAT or LISP  for some applications.

Something  should probably go into some FAQ at some point.....

"Q:   IPv6 could still run out of addresses / Waste  will kill ipv6 too  /
Etc."

"A:   No,  although there is an occasional point of confusion regarding
IPv6 that we
will still run out of addresses:  that is a highly-improbable event:
please see list archives.

>From evaluation of the arithmetic, there is not a reasonable forecast model
that can be
made that would start from realistic assumptions  about network growth and
could come to the conclusion that depletion of IPv6  would be a possibility
under  current regional registry allocation policies based on justified
need,
even allocating up to a couple  dedicated  /48s  per  person  up to the
expected maximum population capacities of earth....


--
-JH

When the IETF decided on 128 bit addresses it was taking into consideration
> /80 sized subnet.  Prior to that it was looking at a 64 bit address size
> and allocating addresses the IPv4 way with lots of variable sized
> networks.  This was changed to /64 subnets to accomodate 64 bit MAC.  After
> that there was discussion about how many subnet should be enough for 99.99%
> of sites which gave /48 per site using /64 sized network.  That
> 281474976710656 sites or 35184372088832 out of the /3 we are currently
> allocating from.
>
> Now there are very few sites that need 65536 subnets and those that do can
> request additional /48’s.
>
> Now if you assume the earth’s population will get to 25B, and every person
> is a site, that still leaves 35159372088832 sites.
> And if each of those people also has a home and a vehicle, that still
> leaves 35109372088832 sites.
>
> Handing out /48’s to homes was never ever going to cause us to run out of
> IPv6 space.  Even if the homes are are connected to multiple providers
> there isn’t a issue.
>
> Mark
>
> > On 21 Dec 2017, at 7:57 am, William Herrin <bill at herrin.us> wrote:
> >
> > On Wed, Dec 20, 2017 at 1:48 PM, Mel Beckman <mel at beckman.org> wrote:
> >
> >> I won’t do the math for you, but you’re circumcising the mosquito here.
> We
> >> didn’t just increase our usable space by 2 orders of magnitude. It’s
> >> increased more than 35 orders of magnitude.
> >>
> >
> > Hi Mel,
> >
> > The gain is just shy of 29 orders of magnitude. 2^128 / 2^32 = 7.9*10^28.
> >
> > There are 2^128 = 3.4*10^38 IPv6 addresses, but that isn't 38 "orders of
> > magnitude." Orders of magnitude describes a difference between one thing
> > and another, in this case the IPv4 and IPv6 address spaces.
> >
> >
> > Using a /64 for P2P links is no problem, really. Worrying about that is
> >> like a scuba diver worrying about how many air molecules are surrounding
> >> the boat on the way out to sea.
> >>
> >
> > It's not a problem, exactly, but it cuts the gain vs. IPv4 from ~29
> orders
> > of magnitude to just 9 orders of magnitude. Your link which needed at
> most
> > 2 bits of IPv4 address space now consumes 64 bits of IPv6 address space.
> >
> > Then we do /48s from which the /64s are assigned and we lose another 3 or
> > so orders of magnitude... Sparsely allocate those /48s for another order
> of
> > magnitude. From sparsely allocated ISP blocks for another order of
> > magnitude. It slips away faster than you might think.
> >
> > Regards,
> > Bill Herrin
> >
> >
> > --
> > William Herrin ................ herrin at dirtside.com  bill at herrin.us
> > Dirtside Systems ......... Web: <http://www.dirtside.com/>
>
> --
> Mark Andrews, ISC
> 1 Seymour St., Dundas Valley, NSW 2117, Australia
> PHONE: +61 2 9871 4742              INTERNET: marka at isc.org
>
>


More information about the NANOG mailing list