[ppml] NANOG IPv4 Exhaustion BoF

[Geoff Huston]

> Currently, there are "transfer" policy proposals in 3 RIRs.  They  
> range from the APNIC proposal which I would characterize as imposing  
> minimal constraint/regulation to the ARIN proposal which I would say  
> has maximal constraint/regulation.

The RIPE policy proposal is very similar to APNIC in terms of the 
constraints applied to the parties of a transfer.

>  From my personal perspective, I  
> would imagine the APNIC model could result in attempts by some folk to  
> "corner the market" while the ARIN model (in particular, the  
> restriction against subdividing) could result in the creation of more  
> liberal alternative registries, further eroding the role ARIN plays in  
> its role within the addressing community.


I have wondered about "cornering the market", and I suspect that there 
are a number of views of the future that may incent certain behaviours.

Some folk may be of the view that IPv4 + NATs is a long term viable 
proposition, and may believe that in such a long term scenario the value 
of IPv4 addresses may rise steadily over time.

Others may be of the view that one of the major elements in an IPv6 
transition is the incentive to stop deploying IPv4 and that may well be 
based in an escalating value of IPv4 addresses, which would increasingly 
provide economic incentives for entities to deploy IPv6 as their 
mainstream technology base with minimal IPv4 translation services.

The diversity of views abot scenarios and timings will drive different 
behaviours - some entities may act on the view that once the shock of 
exhaustion wears off and the related market price has factored in 
scarcity premiums as well as the initial shock, the drive to IPv6 will 
assume a new level of momentum, and the IPv4 demand will drop, and with 
declining demand so will market prices - i.e. some folk would see no 
value in hoarding and believe that maximal value will be seen in the 
initial operation of short term market that will be overtaken by IPv6 
relatively quickly. Other entities may be of the view that the inertia 
of this industry is sufficiently large, and the capability of dense NAT 
deployments sufficiently tenable that an IPv4 market may be used for a 
long period. If this is accompanied by a view that demand will continue 
to outpace supply then an escalating price is an obvious outcome.

Personally, I would tend to the first view - that escalating price in an 
Ipv4 address market would rapidly drive the industry into IPv6. This 
tends to gather its own momentum given the observation that while 
initial Ipv6 deployments have the added cost imposition of also having 
to provide mechanisms to access IPv4 networks, as the mass of the Ipv6 
deployment increases there is a natural cost transfer and it would be 
likely that we see a tipping point whereby Ipv6 deployments could be 
undertaken with no legacy access provisions, thereby effectively 
transferring the burden of translation to the legacy IPv4 network.

In either case the major benefit of this approach of leaving this to the 
actions of the market itself is that it does not require oracle-like 
behavior of regulators or even industry players - rather than attempting 
to drive the industry to certain outcomes, with an historical success 
rate that one could only describe as stunningly poor (remember GOSIP, 
remember X.25, remember even QoS...) such an approach of exposing 
potential suppliers and recipients to each other in the context of an 
open competitive market place essentially allows each actor to make 
their own decisions as to technology choices based on their own guesses 
about the future.



> Much of the discussion to date has been the equivalent of rearranging  
> deck chairs on the Titanic.  IPv4 is sinking.  Fast.  You can create a  
> fractal maze in front of people trying to get to lifeboats (the ARIN  
> model) or you can facilitate a stampede (the APNIC model).


Firstly I have to say that the combination of references to fractal 
mazes, lifeboats and stampedes all crammed into one sentence is an 
achievement all of its own. Well done! :-)

I'd hardly characterize the APNIC policy proposal in such dramatic 
terms. The APNIC model as it stands leaves the operation of any 
associated market to the industry players themselves. If such a market 
forces prices up to astronomical heights then the incentive to head 
immediately to IPv6 becomes all the greater, and the prospects of any 
Ipv4 address market quickly losing relevance also becomes greater, 
simply because the industry has been forced to undertake a transition 
that it has been deferring previously. For many such a long anticipated 
outcome of actual transition rather than oblique lip service to a token 
IPv6 effort is not exactly a disaster. Some would say its a fine outcome 
as long as you're not part of any collateral damage in the associated 
blame game in the aftermath. So its not as if there is absolutely no 
alternative here. And its not as if such forms of market outcomes spell 
doom and disaster of an out of control stampede. This is not quite the 
wild days of 2000 and the GSM auction debacle appears to have hosed down 
some of the more insane players in this industry. This is a relatively 
low margin commodity business down at the plumbing side and it may well 
be that market pricing may well be influenced by the limited amounts of 
capital that are within the industry these days.

One could say that the ARIN model attempts to constrain demand, and in 
so doing one possible interpretation is that such a degree of externally 
imposed constraint effectively lengthens the pain of attempting to 
stretch IPv4 well beyond its originally assumed deployment scope. This 
may not be the wisest course of action if you want to provide some form 
of natural incentives to get over this somewhat strange and distorted 
scenario of IPv4 address exhaustion tht we are facing.


>  If the  
> goal is to get folks to use IPv6, the APNIC model has certain  
> advantages.  If the goal is to encourage the creation of alternative  
> registries, the ARIN model has certain advantages.
> 
> However, it should be noted that in either case, the core aspect of  
> the end result is the same: people are going to be fighting for seats  
> on lifeboats and the ship will be on the bottom of the ocean.


Hmm - lets see, on this topic we've now gone through airplane crashes, 
steam train wrecks, stampedes and now sinking ships. What next? Meteor 
strike? :-)

     Geoff