ULA (was Re: ARIN Board Advises Internet Community on Migration to IPv6)

[Shane Kerr <shane@time-travellers.org>]

All,

On Mon, May 21, 2007 at 12:35:56PM -0700, Fred Baker wrote:
> If the desire is for a space that can be used within an ISP network
> and is guaranteed to not overlap that of any routing peer,  placing
> the 32 bit AS number into the upper bits of the "Global ID"  gives
> each provider a /40 to play with, with no effort on the IANA's
> part. Why not just do so?

Using a 32-bit AS number as the basis of an address makes a lot of
sense. It doesn't really solve the "missing RFC 1918" problem for IPv6
though. Anyone who can get an AS number can get a /32 from an RIR, so
there's not much motivation for this type of private space. (Note, I'd
fully support a draft that outlines the 32-bit AS address space idea.)


There are already at least two ways to get unique local IPv6 blocks
today:

1. get a /48 from someone with an RIR allocated block (of course)

2. take an IPv4 address and convert it to a 2002::/48 block

In the case of #1, you can either use your own block, or get some
sort of guarantee from the organization you got it from that they
won't hand it out to someone else.

And you *can* use the pseudo-random ULA from RFC 4193, which anyone
who understands probability will be quite okay with. (The chances that
*you* will ever connect to another network that chose the same ULA are
really, really small. In fact the chances are really small that
*anyone* will ever have a collision are very small, if we assume that
these networks are not highly connected.)


I really think ULA is a solution looking for a problem.

--
Shane