[182019] in North American Network Operators' Group
RE: Dual stack IPv6 for IPv4 depletion
daemon@ATHENA.MIT.EDU (Matthew Huff)
Thu Jul 9 12:58:16 2015
X-Original-To: nanog@nanog.org
From: Matthew Huff <mhuff@ox.com>
To: Owen DeLong <owen@delong.com>
Date: Thu, 9 Jul 2015 16:53:36 +0000
In-Reply-To: <62F58572-FA6C-4CB1-89EE-5C52B428D93F@delong.com>
Cc: NANOG list <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Sure, they may be 100,000+ networks like that in non-technical households.=
Maybe. I doubt it, but still that would be like 0.01%. Many consumer syste=
ms have trouble with L3 hops (mDNS/Bonjour, etc...). First thing tech suppo=
rt will suggest it to put them on the same network. People have been taught=
this. My experience is that most people that even have a second network, i=
t's their AP that sets up a Guest network, and even then, it doesn't route =
between the networks (that's sort of the whole idea).
If an ISP wants to give out a /48, great for them. If they want to give out=
only a /56, I say that's fine. What's more important to me is that they im=
plement IPv6. Arguing about prefix size and SLAAC vs DHCP rather than just =
go ahead and implement things, to me is just silly. When IP was first deplo=
yed, we didn't have DHCP (bootp was still in it's infancy), no mDNS, etc...=
Lots of things grew up after the fact. I agree that we can't foresee what w=
ill happen in the future, but that to me just proves my point. Worrying abo=
ut the ability to create complex topologies in home networks that may or ma=
y not ever be needed or wanted just seems absurd to me.
----
Matthew Huff=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 | 1 Manhattanville Rd
Director of Operations=A0=A0=A0| Purchase, NY 10577
OTA Management LLC=A0=A0=A0=A0=A0=A0 | Phone: 914-460-4039
aim: matthewbhuff=A0=A0=A0=A0=A0=A0=A0 | Fax:=A0=A0 914-694-5669
-----Original Message-----
From: Owen DeLong [mailto:owen@delong.com]=20
Sent: Thursday, July 9, 2015 12:36 PM
To: Matthew Huff
Cc: Marco Teixeira; Harald Koch; NANOG list
Subject: Re: Dual stack IPv6 for IPv4 depletion
> On Jul 9, 2015, at 08:42 , Matthew Huff <mhuff@ox.com> wrote:
>=20
> What am I missing? Is it just the splitting on the sextet boundary that i=
s an issue, or do people think people really need 64k subnets per household=
?
>=20
It's the need for a large enough bitfield to do more flexible things with a=
uto-delegation in a dynamic self-organizing topology.
8 is 2x2x2 and there's really no other way you can break it down. (2x4, 4x2=
, 2x2x2 is it.)
16 is 2x2x2x2 and allows many more possible topologies (4x4, 2x4x2, 2x2x4, =
2x8, 8x2, etc.)
> With /56 you are giving each residential customer:
>=20
> 256 subnets x 18,446,744,073,709,551,616 hosts per subnet.
The host count is irrelevant to the discussion.
>=20
> I would expect at least 95.0% of residential customers are using 1 subnet=
, and 99.9% are using less than 4. I can understand people complaining when=
some ISPs were deciding to only give out a /64, but even with new ideas, n=
ew protocols and new applications, do people really think residential custo=
mers will need more than 256 subnets? When such a magical new system is dev=
eloped, and people start to want it, can't ISPs start new /48 delegations? =
Since DHCP-PD and their infrastructure will already be setup for /56, it ma=
y not be easy, but it shouldn't be that difficult.
I would expect that basing decisions about limits on tomorrows network on t=
he inadequacy of today's solutions is unlikely to yield good results.
Further, I'm not so sure you are right in your belief. I suspect that there=
are many more networks in most households that you are not counting. Sure,=
those networks are currently usually disjoint, but do you really think it =
will always be that way in the future?
Every phone is a router. Ever tablet is a router. Cars are becoming routers=
and in some cases, collections of routers. Set top boxes are becoming rout=
ers.
Utility meters are becoming routers.
Laptops and desktops are capable of being routers.
> I know the saying "build it and they will come....", but seriously....
>=20
> I'd rather ISPs stop discussing deploying IPv6, and start doing it.
I'm all for that, but do you have a valid reason not to give out /48s per e=
nd site? Just because /56 might be enough doesn't cut it. I'm asking if you=
can point to any tangible benefit obtained from handing out /56s instead? =
Is there any problem solved, labor saved, or any other benefit whatsoever t=
o giving out /56s instead of /48s?
If not, then let's hand out /48s until we discover one.
Owen
