[ppml] Proposed Policy: IPv4 Countdown

[Tony Hain]

michael.dillon at bt.com wrote:
> > I would like to see some discussion on this, even though I comprehend
> > the reasons given below for the rejection and acknowledge that they
> > are valid.  My motivation for speaking up is to see if there might be
> > a way that the spirit of the proposal can be pushed forward in ARIN
> > even if the particular proposal has mechanics that are problematic.
> 
> ARIN has long had a problem with ill-thought-out, hastily written policy
> proposals that were shoved into the public arena before they were ready.
> In one sense, this is yet another of these. Like other poorly crafted
> proposals it also suffers from the "plain English" problem. What is an
> A-Date or T-Date? Why  can't the authors just say what they mean in
> plain English? How many times did you have to read the proposal, jumping
> back and forth in the text, to figure out just what they are saying.
> This is bad. The Rationale section is there to allow authors to explain
> the reasoning behind a policy, not to explain the meaning of the policy
> text.

First you have to give them credit for writing in English at all, then it
would be useful to realize that they were likely trying to be as precise as
possible by using a mathematical approach to the description. "Plain
English" does not always translate well, even among those who believe they
are native English speakers.

> 
> In addition, the authors suggest that ARIN policy should require ARIN to
> take some action when IANA resources reach a certain level. But IANA is
> not controlled by ARIN. Does IANA even report on the size of their pool
> in /8 equivalents? How is this to be measured?

Yes, IANA maintains a public document of their allocations:
http://www.iana.org/assignments/ipv4-address-space
to save you the trouble of parsing it, there is a graph at:
http://www.tndh.net/~tony/ietf/IANA-IPv4-distribution.pdf
Similar graphs are published by ARIN, APnic, RIPE-ncc from time to time. You
are likely to find +/- 1 or 2 in various buckets because the records are
inconsistent for some of the pre-'95 allocations.

> 
> But that is not all that is undefined. Does ARIN have a clear policy
> definition of "critical infrastructure" as referred in the proposal? The
> proposal refers to "projections" but does ARIN actually make such
> projections?
> 
> And of course, the policy requires ARIN to terminate allocating IPv4
> ranges which is entirely contrary to ARIN's charter. ARIN exists to
> allocate IPv4 addresses and the only reason for this to stop is for the
> supply to be exhausted. As long as IANA gives ARIN IPv4 addresses, ARIN
> should continue to allocate them according to its policy guidelines.

This is what will happen in practice at all the RIR's. 

> 
> In case you hadn't noticed, this policy proposal was made by people from
> outside the ARIN region. The same proposal was put before APNIC in their
> own region. They are attempting to create a global policy without
> following the global policy process of NRO
> http://www.nro.net/policy/index.html

The reality is that there will never be global agreement on how to manage
down the tail end of the pool, because there will never be global agreement
on the fundamentals that are used to enforce allocation policy, despite the
existence of the NRO. The proof of that viewpoint is the very existence of
the RIR's themselves. If there could be global agreement on policy, we would
only be dealing with IANA as the central body for managing the resource.


> 
> Note that in the NRO process, the global policy is ratified by the ICANN
> board of directors before coming into action. Also note that IANA is one
> of the functions of ICANN, in other words if you want to make policies
> based on IANA resources or IANA actions, you should go through the NRO.
> Two of the three existing global policies actually do deal with IANA so
> presumably the process works reasonably well.

The NRO is really all about how to get the RIRs to play nice in the sandbox
with each other. Yes there is a formal process, but at the end of the day
the process that matters is on the outbound side of the RIR to their
members. 

> 
> As for anti-trust, well, this policy sounds like somebody making up
> rules just for the sake of making up rules. The real goal seems to be
> publicity of the fact that we are now in the wind-down phase of IPv4 and
> IANA could run out of free /8s as early as 3 years from now. In my
> opinion, ARIN policy is not the way to solve a publicity problem and not
> the way to solve an education problem.

I agree that RIR policy is not a place to be solving publicity problems. At
the same time the members need to be aware of the pending reality, and if
there are going to be changes needed for the post exhaustion event, that
planning needs to start well in advance. While Geoff currently shows a 2011
date, my projection has been holding steady at the end of 2009 for the end
of the RIR pools:
http://www.tndh.net/~tony/ietf/IPv4-pool-combined-view.pdf
For those that might remember my initial date as 2008, that was for the IANA
pool. There was a marked decrease in the rate of allocations from IANA to
RIPE & APnic last year so that flattened out the IANA curve, but the
outbound side of the RIRs to their membership has not shown any substantial
variation. In any case Geoff has been looking at what will be needed from an
RIR perspective to support the post exhaustion environment. 

An observation I have been working from lately : It is simple human nature
to ignore the problem until it becomes a crisis. Given that, most Network
Managers will not act until they run into a problem getting IPv4 space. 

Despite the best intentions in RIR policy management, we will burn through
the remaining IPv4 space sometime in the next 2-4 years, and IPv4 addresses
will become a commodity traded on E-bay. It is simply too late to make any
difference by modifying RIR policy. That said, I actually expect the 'sale'
activity to be short lived, because venture capital will quickly collect the
available resource and return it to market in 'lease' form. Today, while the
dogma proclaims that addresses are not property, people routinely get away
with 'leasing' a single publicly routable IPv4 address for ~$5/day (hotel
VPN fee), and the average global 'lease' rate I hear for static publicly
routable IPv4 on long term contract is ~$1/day/address (YMMV). Once the
taboo is totally trashed by the reality of global market trading in the
finite resource, those prices will only go up. Yes the unused existing
allocations will flood into the market, but that surge will only lower the
investment price for those with longer term vision that see the real value
in leasing a commodity that there will not be any more of.

The other thing to note is that despite whatever space might suddenly appear
as unused and available, the existing consumption rate is over 1 /8 per
month and growing. Note that ongoing consumption is occurring globally:
http://www.tndh.net/~tony/ietf/IPv4-delegated-per-RIR.pdf
It will take a lot of reclamation, for a very long period of time to avoid
having the price skyrocket for adding new customers or services. ISPs will
have to pay whatever the market demands until their -customers- stop using
IPv4. Of course they will pass that added cost along, so at contract renewal
time the customers will be astounded at the rate increases just to keep
doing what they had been for years. Eventually this will resolve itself as
people realize the machines they have are capable of IPv6, some services
will offer IPv6 access forcing others to follow suit or loose eyeballs, at
which point the only ones that will even be aware of IPv4 or care are those
that just refuse to evolve.

Tony