[ppml] Policy Proposal 2004-3: Global Addresses for Private N etwork Inter-Connectivity

[Michael.Dillon at radianz.com]

>I wish to reiterate this call for comments. 

I reiterate Bill's reiteration.

> RFC-1918 and RFC-2050 are somewhat ambiguous concerning the use of 
> registered IP addresses when connecting separate private networks. 

In particular, I disagree with the basic premise of this 
proposed policy. I think that RFC-2050 is very clear:

      a)  the organization has no intention of connecting to
          the Internet-either now or in the future-but it still
          requires a globally unique IP address.  The organization
          should consider using reserved addresses from RFC1918.
          If it is determined this is not possible, they can be
          issued unique (if not Internet routable) IP addresses.

It does not mandate following RFC 1918 and it does not
require any specific reasons for determining that RFC 1918
usage is not possible.

Note that RFC 1918 has the following text in it:

    Category 3: hosts that need network layer access outside the
                enterprise (provided via IP connectivity); hosts in
                the last category require IP addresses that are
                globally unambiguous.

It seems perfectly clear to me. When you interconnect two
private RFC 1918 networks you should use globally unique
IP addresses to do so. And this is sufficient justification
for an allocation from a registry.

I would like to see a comment from ARIN staff as to whether
or not they interpret the current rules this way.

Note that I work for a company which operates a global IP
network in 20 countries using over half a million globally
unique registered IP addresses. Our network is not connected
to the network and we have no intention of ever connecting
to the Internet. Our business model is based on interconnecting
the private networks of businesses in the financial services
industry so that they can provide services to one another.
This means that we primarily interconnect networks that use
RFC 1918 addressing internally.

So the bottom line is that I understand current policy to say
that any infrastructure which interconnects two or more
networks, private or otherwise, should use globally unique
IP addresses and that this infrastructure can justify their
need for addresses in the normal way.

Now I do agree that ARIN policies and ARIN forms are
becoming dreadfully obsolete, filled with anachronisms
and failing to meet the new realities of the IP
internetworking world. I'm not at all surprised that
people are confused by this and would like to clear
up some corners of ambiguity. If this policy only included
the first sentence then I would wholeheartedly support
it. But right now, it seems to be based on misunderstanding
and, possibly, on ARIN mistakes in applying the existing
policy. I'd really like to get to the bottom of this.

Michael Dillon