[arin-ppml] DRAFT POLICY ARIN-2011-1: GLOBALLY COORDINATEDTRANSFER POLICY (Legecy space)

[Milton L Mueller]

Owen, 
it's nice to have a more detailed discussion of v4-v6 migration economics, because while I see lots of IPv6 evangelism I don't see much serious discussion of how it is really going to happen. Responses below

> -----Original Message-----
> 
> Let me attempt to explain it a  bit more accurately:
> 
> The Mueller theory:
> --------------------------
> At T1, upgrading network to IPv6 costs $1000, more ipv4 addresses cost
> $1000. However, more IPv4 addresses will allow T1's users to reach the 
> still-only-on-IPv4 parts of the internet while upgrading to IPv6 without 
> IPv4 addresses will leave T1's users without such access, or, with severely 
> degraded access to those parts of the internet that are still IPv4-only. 
> Rational actor picks buying more v4.

OK, that makes sense and is not much different from what I said.

> At T2, upgrading network to IPv6 costs $1000, more ipv4 addresses cost
> $2000. Same constraints on the usefulness of the IPv4 addresses apply. 
> "Rational" actor ignores this disparity of usefulness and chooses the cheap
> connectivity. T2 loses his customers as a result and goes bankrupt.
> The price of IPv4 addresses drops as T2s IP addresses are now available 
> cheap in a bankruptcy fire sale.

Three observations. 
First, we have some current experience with a bankruptcy sale of IPv4 addresses and it wasn't that cheap. I doubt that it will reduce prices in the short term. But that's a tangential point.
Second, you seem to be saying that anyone who opts for IPv6 in the current environment without also adding to their IPv4 stock will go bankrupt because of the incompatibility.
That's also in agreement with what I've been saying about the dual stack migration strategy, and it's why I believe the transition will take longer than most people think and why we need to pay a lot of attention to v4 transfer/trading policies. 
Third, the actual rational strategy here would be to try to spend money on upgrading to IPv6 while also acquiring more v4 addresses, but to deploy methods, such as CGN or DS-Lite, that reduce the number of new IPv4's one needs. 

> The DeLong theory:
> --------------------------
> At T1, upgrading network to IPv6 costs $1000, more ipv4 addresses cost
> $1000. However, more IPv4 addresses will allow T1's users to reach the 
> still-only-on-IPv4 parts of the internet while upgrading to IPv6 without IPv4 
> addresses will leave T1's users without such access, or, with severely degraded
> access to those parts of the internet that are still IPv4-only. Rational actor 
> picks buying more v4. Further, since T1 sees the writing on the walls, he also 
> spends $1,000 to upgrade his network to IPv4/IPv6 dual-stack so that his
> customers can reach both the IPv4 and the IPv6 capable parts of the internet.

Yes. No difference between us here, really. In essence, all you are telling us is that a rational actor will have to spend to do both: upgrade to IPv6 for the long term and get more IPv4 in the short term. However, it does not undercut the fact that a rising price for v4s will increase the likelihood of stragglers taking the need for a v6 migraiton more seriously.

> At T2, upgrading network to IPv6 costs $1000, more ipv4 addresses cost
> $2000. Same constraints on the usefulness of the IPv4 addresses apply. 
> "Rational" actor recognizes that keeping his customer has a value in excess of
>the price disparity and purchases the more expensive IPv4 addresses. 
> Unfortunately, this unexpected additional cost of IPv4 addresses means that 
> the $1000 he was planning on spending to get to IPv6 has now been
> diverted into the additional cost of IPv4 and there is no money to
> upgrade to IPv6. The provider is now trapped	in a very difficult position 
> where it will become ever-increasingly expensive to grow his network with IPv4,
>but he desperately needs capital to deploy IPv6.
> 
> Does this make sense to anyone?

Sure, it makes sense, although as I said before the most likely outcome of the scenario you present is that people will deploy NAT or things like Dual Stack Lite so that their consumption of new IPv4 addresses will be reduced as they (v4 addresses) become more expensive. 

The gist of your scenario is that the migration to IPv6 will be expensive; that IPv6 is a net ADDITION to cost, and that no ISP can afford to abandon IPv4 any time soon. Thanks for joining me in promoting this message. 

What I don't understand is why we are (seemingly) having an argument about policy.  My policy take-home is that ARIN and the RIRs need to do everything they can to make the transfer of IPv4s from less valuable, less efficient uses to more valuable uses as quick and easy as possible. My take home is that people like you need to stop whinging about the monetization of addresses and recognize it as a fact and figure out how to deal with it. All of the scenarios presented above, whether the Mueller or DeLong version, lead to that conclusion. 

Or maybe there is something I am missing. Is there some way that a policy of resisting and discouraging IPv4 markets will: 
a) eliminate the need for growing ISPs to acquire IPv4 blocks? 
b) reduce the cost of making the investments required to upgrade to IPv6; 
c) prevent the supply of IPv4 blocks from decreasing? 

If your answer to all those questions is NO, and I think it is, you need to step back and ask yourself why what we are allegedly disagreeing about. 

--MM