[AusNOG] RFC7278 - "Extending an IPv6 /64 Prefix from a Third Generation Partnership Project (3GPP) Mobile Interface to a LAN Link"

[Karl Auer]

On Wed, 2014-07-02 at 21:31 +1000, Joseph Goldman wrote:
>   Going a bit off-topic, towards IPv6 in general as I'm still catching 
> up on the standards of use for IPv6, but I am yet to understand the 
> reason for recommendations to give such large blocks to customers?

Since you posted to the list I'm going to stick my oar in, while also
looking forward to Mark's reply.

The basic rule with IPv6 addressing is "forget addresses - count
subnets".

The number of addresses per subnet is ridiculously high; that's
intentional. It doesn't matter how many hosts you have on a subnet, you
will run into other physical limits before you run out of addresses. You
have to have two hundred million billion hosts in a /64 subnet before
you reach striking distance from single-digit percentage utilisation.
Any reasonable number of hosts we can think of, in fact any UNreasonable
number of hosts we can think of, even when we double and redouble our
wildest imaginings, is still basically zero percent usage to many
decimal places. This is great - no more slicing and dicing, no more
rearranging things to fit, all your subnets are the same size, life is
easy and good.

So leaving addresses behind, how many subnets should a subscriber get?
The first cut was "65536" - everyone should get a /48. That's been
relaxed a bit (a mistake IMHO) and the current RFCs allow smaller
assignments to end users, though the RIRs still allocate nothing smaller
than a /48, and the IPv6 world seems to be converging on /48 as the
longest globally routable prefix (roughly equivalent to a /24 in the
IPv4 world).

Internode hands out /56 prefixes - that's 256 subnets. They chewed on
the decision for a while - /60 was considered (16 subnets) and rejected,
but it seems /48 confounded the imagination even of this usually fairly
visionary company, and they ended up going with /56.

The logic behind "bigger is better" is very simple - you never want a
customer to come back to you for more address space. It requires
resources to satisfy such a request. From a customer perspective, you
never want to have to go back for more, either. Especially if the new,
larger allocation means renumbering. 256 subnets looks like a lot today,
but who knows what the future will bring?

>   You talk about a /64 being handed out to customers, even this I found 
> exceptionally large for a home, which even with smart devices becoming 
> the norm would you say its likely to reach 100 needed IP's? let alone 
> thousands?

Again, it's not addresses, it's subnets. That /64 is ONE subnet. Can't
even have a guest wifi with that! Can't separate the home network from
my work network. Can't have a test network, or a gaming network. It will
only get worse - in the future the home entertainment system will want a
few subnets, the kitchen will want a few subnets, the home automation
system will need a few subnets, sons and daughters will want a few
subnets each for their gaming systems and the like, obviously guests
will want a few subnets to themselves... pretty soon even a /56 with
only 256 subnets is looking pretty skimpy.

The counter argument to all this is that if you are profligate enough,
you can reduce even the most plentiful resource to nothing pretty
quickly. There is some truth in that. However, if you do do the maths
you quickly realise what a staggeringly large amount of space there is
to burn through. And is it really waste? I define waste as using a
resource to no good effect, but in this case the resource is being used
with several very good effects - homogeneity, predictability, no more
slicing and dicing, lower support costs, reduced human error etc etc.
That's a direct monetary benefit, and that's just on the network admin
side.

IPv6 opens up a world of plentiful address space for use by everybody.
We may not yet see how that abundance will be used, much as Bill Gates
couldn't imagine how anyone could use more than 640K or RAM... but we
should not bolt the door for others just because our limited vision sees
no reason to go outside. There is simply no point to hoarding IPv6
address space. There is literally more than enough for everybody.

Regards, K.

PS: This is a discussion that pops up regularly. It's can get a bit
religious. Be warned :-)

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Karl Auer (kauer at biplane.com.au)
http://www.biplane.com.au/kauer
http://twitter.com/kauer389

GPG fingerprint: EC67 61E2 C2F6 EB55 884B E129 072B 0AF0 72AA 9882
Old fingerprint: B862 FB15 FE96 4961 BC62 1A40 6239 1208 9865 5F9A