[ppml] Resurrecting ULA Central [was: Re: Policy Proposal 2006-2: Micro-allocations for Internal Infrastructure - to be revised ]

[Thomas Narten]

David Williamson <dlw+arin at tellme.com> writes:

> On Thu, Apr 20, 2006 at 04:36:37PM -0400, Thomas Narten wrote:
> > Would there be interest in resurrecting the ula-central document?

> Speaking only for me, yes.

> It would seem mildly clever to put the AS number somewhere in the
> network number (pick an appropriate range of bits in the prefix), and
> then hand out the appropriate /48 implicitly with any AS received from
> an RIR (including legacy AS holders, obviously).

OK, I'll bite.

The original ULA Central approach just called for a single way of
getting ULA assignments. I.e., just one "registry". The registry would
be responsible for assigning unique values for the "global ID" field,
which is 40 bits. Discussions at the time assumed/implied that the
RIRs would play this role.

During the discussion last week, it occurred to me that it might make
sense to actually have a small number of different kinds of "ula
central" addresses (each still globally unique and guaranteed not to
collide with those allocated by different means). That is, have the
"global ID" field consist of a (say?) 4 bit "address type", followed
by a (say?) 36 bit "global ID" field. One of the "address types" could
indicate (as you suggest) "ASN number", in which case the "global ID"
field would just contain an ASN number. Voila. All ISPs already have
one.

Same could be done for end sites, for which IANA already maintains a
registry for PENs (Private Enterprise Numbers). Some 25K of them have
already been assigned, and they are trivial to obtain. Voila, we've
already grandfathered in a huge number of end sites.

(I'm not sure now how much the above was ever discussed in the
discussions leading up to the ULA central document, but I'd be in
favor of considering such an approach.)

Thomas