[ppml] ARIN Policy Proposal 2002-9

David Conrad david.conrad at nominum.com
Wed Oct 2 14:02:28 EDT 2002


Again, speaking personally, not as an ARIN board member...

On 10/2/02 9:47 AM, "George Cottay" <cottay at qconline.com> wrote:
> I'm confused by discussion here about needs for non-routed IP's other than the
> present 10, 172, and 192 space already reserved.  Especially given the size of
> the, I cannot for the life of me imagine an organization needing
> more. Even if one were to divide on the basis of the old class C, that leaves
> upwards of 65,000 possible subnets with which to play.

In a previous life (in the very early days of APNIC), I was the (equivalent
of the) IP analyst at APNIC (as well as CEO, CFO, programmer, sys admin, and
janitor :-)).  I'd get two or three requests a month from organizations that
wanted address space but who claimed they had no intention of ever
connecting that address space to the Internet.  When I asked why they
couldn't use the RFC 1918 space, the requestors invariably told me they
didn't want to have to renumber should they acquire another company/merge
internal groups/etc.

Don't know if ARIN IP analysts get these sort of requests, but it wasn't
that uncommon in the AP region in the mid- to late-90's.  As an aside, I
will note that (last I checked) all of the organizations who claimed to me
(back in the mid '90s) they'd never connect to the Internet are, in fact,
connected to the Internet today.

> I'm even more confused by mention of a need for public addresses that are not
> routed.  I thought routing was the most significant difference between public
> and private space.

Uniqueness is the only attribute the registries can provide that separate
the addresses they allocate from any randomly generated 32 (or 128) bit
integer.  That uniqueness is often of value independent of whether the
address space is to be used on the Internet or not.  I personally know of
several very large organizations that are using the exact same 10/8 prefixes
internally (perhaps an interesting study for CAIDA or whoever -- of the
folks who use net 10 (et al), how many start with 0 and go up in subnets and
how many assign subnets randomly).  I also have seen the implications of
renumbering when two large organizations using the same prefixes try to

Please note that I am not saying the allocation of unique address space not
intended for use on the Internet is a good or bad idea, just trying to
answer the question.  I will, however, admit that I'm a bit skeptical that
renumbering a private /24 is much of a hardship...


More information about the ARIN-PPML mailing list