[arin-ppml] Policy Proposal 119: Globally Coordinated Transfer Policy

[Leo Bicknell]

In a message written on Mon, Oct 11, 2010 at 01:20:51PM -0400, ARIN wrote:
> Policy statement: Any RIR's member may transfer IPv4 addresses to the
> member of another RIR as long as the two RIRs agree and exercise
> Internet stewardship and the values expressed in RFC2050.

I applaud the attempt at super-simple policy.  It is refreshing.

That said, I wish to play a sort of devils advocate for a moment.

There are activities one RIR's community thinks are "bad", but that
other RIR's communities seem to think are "good".  To some extent
that is why we have separate RIR's, but there is also a desire to
limit the badness globally.  I believe the references to RFC2050
are an attempt to restate such a global limit.

Unfortunately RFC2050 is already long in the tooth, and with the
impending IPv4 exhaustion will become less relevant very quickly.
In particular, I think if you look at a post run-out world, it is
likely different people will come to the same conclusion about the
text in various areas.

To that end, I'm actually not sure what the effect of this policy
would be down the road, and being unable to evaluate that effect I
find I cannot support it.

If you don't care about the details, stop reading now, if you do,
continue.

Let's fast forward 2-3 years from now to a world where IANA and all 5
RIR's are "out" for all practical purposes, and look at some of the
statements in RFC 2050 from that context.

Right from the start, there is something quite interesting:

   By approving this document as a Best Current Practice,the IESG
   asserts its belief that this policy described herein is an accurate
   representation of the current practice of the IP address registries
   with respect to address assignment.  This does not constitute
   endorsement or recommendation of this policy by the IESG. The IESG
   will reevaluate its approval of this document in December 1997 taking
   into consideration the results of the discussions that will be take
   place in the IRE Working Group between now and then.

The document is not a standards track RFC, it is a documentation of the
Best Current Practice /in 1996/.  Indeed, I actually think it's well
overdue to be updated to current practice, but that's another discussion
for another day.

  1) Conservation: Fair distribution of globally unique Internet address
     space according to the operational needs of the end-users and Internet
     Service Providers operating networks using this address space.
     Prevention of stockpiling in order to maximize the lifetime of the
     Internet address space.

"Fair distribution" is a very simple thing when there are plenty of
addresses, we can use the Chinese buffet model (take all you want, eat
all you take).  When there is no more space, and we're looking at a
transfer market (paid or otherwise) it takes on an entirely different
color.  Do we have to ensure everyone gets some, regardless of ability
to pay?  Is distribution based on who has the most money fair?  Does
fairness require the RIR's to more aggressively revoke space from those
who are not using it to fairly distribute to those who need it?  The
last sentence about stockpiling would seem to indicate that is
important.

      (Under the end user "Assignment Framework" section)

      c)  the organization's actual requirement for IP space is
          very large, for example, the network prefix required to
          cover the request is of length /18 or shorter.

All of the RIR's have changed their policy to allow longer than /18
prefixes to end users, and I think off the top of my head many now allow
longer prefixes to end users than ISP's.  As we get closer to
exhaustion, prefix length will get smaller.

   Organizations will be assigned address space based on immediate
   utilization plus 1 year projected utilization.  A prefix longer than
   /24 may be issued if deemed appropriate.  Organizations with less
   than 128 hosts will not be issued an IP address directly from the
   IRs.  Organizations may be issued a prefix longer than /24 if the
   organization can provide documentation from a registry recognized ISP
   indicating the ISP will accept the long prefix for injection into the
   global routing system.

I believe there are already web-hosters who have less than "128
hosts", but have received space because they host hundreds of web
sites, each of which might need an IP address.

I'm not pointing these out to argue about any of them individually,
but simply to point out the document, like any, becomes less relevant
over time.  This makes sense, as it is a best current practice, and
not the hard and fast requirements many seem to want to make it.

We're now in a post run out world, and want to assess "need".  There
will be thousands of folks who can demonstrate technical need the
way we have always done it, but we can't service them all.  So a
new criteria needs to be added into the mix.  All of the RIR's seem
to be moving to money via some mechanism for paid transfers.
Economists and capitalists would love this, since clearly those
with the greatest need for a resource are the ones willing to pay
the most money.

It is easy to see an argument being made as this backlog of need
increases that much of (but not all) our current work to verify
need is unnecessary.  Why spend hours verifying an organization
"needs" a /16 and checking all of their engineering plans when all
they can buy in the market place is a /24?  And if they are willing
to pay $50,000 for that /24, isn't that a better indicator of how
much they need it than any engineering plan?

I'm not sure an RIR could argue a pure dollar based method passes
the smell test, for instance 2050 does suggest "Prevention of
stockpiling in order to maximize the lifetime of the Internet address
space.", so perhaps they have to insure someone isn't just buying
it up to stockpile.  However in such a scheme it would be easy to
say your engineering plans do not need to be reviewed unless you
have paid for more than a /20, because if you have very small amounts
of space its rather hard to call it stockpiling.

In short, I fear the best practice that is 2050 is far shaker ground
going forward than many realize.  It won't ever be tested in a
court, but it will be continuously tested in the court of public
opinion.  It is though a snapshot in time of a continuously evolving
process where we have already seen fit to change its guidance.  I
can easily see a future where four of the five RIR's decide a practice
is the best current practice in the future, but one of the four is
a hold out for whatever reason.  To the extent this is a global policy,
IANA would likely side with the four, noting the represent community
consensus.

While IPv4 is getting all the attention at the current time, IPv6
is, as usual, the more interesting topic.  I urge you to consider
a world 20 years from now, where IPv4 has been removed from all of
the large Internet backbones and we're operating in a primarily
IPv6 only world.  Now re-read 2050.  Does it make sense?  How much
of it matches the best current practice in this future world?

-- 
       Leo Bicknell - bicknell at ufp.org - CCIE 3440
        PGP keys at http://www.ufp.org/~bicknell/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 826 bytes
Desc: not available
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20101013/2662deeb/attachment.sig>