[arin-ppml] ARIN-prop-151 Limiting needs requirements for IPv4transfers

Owen DeLong owen at delong.com
Wed Nov 23 13:53:45 EST 2011

On Nov 23, 2011, at 10:12 AM, Mike Burns wrote:

> Hi Owen, responses inline.
>>> Second, it caps needs-free transfers at the size of a /12, which Owen helpfully points out is just 1/4096th of unicast space. With a cap like this in place, cornering the market is not feasible and total potential profits are limited, making this kind of speculation unsound because the high risk of speculating in this market can not be outweighed by oversized returns.
>> Capping needs-free transfers at a /12 is absurd. That's virtually equivalent to no cap whatsoever. If you want me to consider the needs-free cap meaningful, then, move it to some more rational place like a /22. At /22, I'd still think this was a bad idea, but significantly less damaging than the current proposal would be.
> I am open to reviewing that number, I don’t really think there will be many large transfers to those who cannot justify them.  The number was initially suggested by David Farmer and I don't think it absurd, rather a good middle which would limit market cornering while allowing most transfers to flow smoothly into Whois.

I think it's useless at best and I think that large transfers without justification to speculators are the inevitable conclusion of this policy. A /12 is an incredible number of addresses. It's more than 1/4000th of the total unicast address space. ARIN has less than 1% of members that qualify for such vast resource holdings through the current justification process. That's hardly a middle... It's the far extreme top.

>>> And Prop-151 seeks to recognize what Owen and others fear, that people will see monetary value in IPv4 addresses, and will act accordingly. That means in addition to greed, IPv4 address value will naturally cause people to use this valuable resource efficiently, to an extent that ARIN justification policies simply can not match.  My own opinion, based on the size of allocated but unrouted address space (less the smidgen used only internally) is that there is a great deal of efficiency still to be wrung out of the pool of prior ARIN allocations, not to mention the pool of legacy space. Whether or not you agree, it is axiomatic that valuable assets are used more efficiently than free ones.
>> Do not attribute your ill-conceived conclusions to me. Nowhere have I stated that value of IP addresses would, in fact, cause them to be used efficiently. In fact, I have argued quite the opposite that value without policy will lead to speculation and artificial pricing not dissimilar from the Tulip Mania of the 1600s. You continue to use this fallacy of "allocated but unrouted" as if route announcements visible to your perspective are somehow an accurate measure of utilization. They are not. There are many addresses in perfectly legitimate use that are not advertised in routing tables visible to you. I won't disagree that more efficiency could be obtained from the pool of prior ARIN allocations, but, I'm not convinced that it is as big a gain as you seem to think. As to your claim that it is axiomatic that valuable assets are used more efficiently than free ones, I am unconvinced. Unless you define efficiency solely in terms of the assets landing in the hands of the highest bidder and not based on them serving the highest societal purpose, your axiom can be shown to break down many times throughout history.
> You misunderstand what I said. What I meant was that  you fear the monetary value of IPv4 addresses will lead to hoarding and speculation, you have clearly stated that before.  My point was that if you recognize that a monetary value on IPv4 addresses could drive behaviors like hoarding and speculation, you must also agree that it will drive behaviors like utilizing them more efficiently and selling un- or under-used space.  Absent any real data, it's only a matter of opinion as to what the difference between allocated and advertised space represents. In my opinion it represents waste for the most part, as the 10/8 block is available for private use. As to your last point, addresses will land in the hands of the highest bidder no matter what we do.

I may not have properly understood your meaning initially, but, I did not misunderstand what you said.

I wouldn't say that I "fear" the monetary value of IPv4 addresses, but, rather that I recognize that absent appropriate policy to the contrary, these things will happen. They have historically happened in every unregulated market of scarce resources, so, they are likely inevitable here absent contrary regulation. As such, I believe it is our duty to impose that regulation to prevent these negative consequences.

Why does the fact that monetary value will drive hoarding and speculation (horribly inefficient uses of IP addresses) have any automatic relationship to a belief that it will drive efficiency?

Finally, some of us have some real data on how at least some of those addresses you don't see in routing tables are being utilized and I can tell you at least this much about some of them:

1. 10/8 doesn't work in all applications. In some cases, private networks talk to other networks even though they don't talk directly to the internet. There is the potential that those networks need addresses which are unique across the other internetworks that they do talk to, some of which may be connected to the global internet. As such, globally unique unicast addresses may be completely justified and necessary in those applications.
2. Many of those networks are as efficiently utilized (80+%) as any policy requirement in the modern ARIN structure.

Just because it isn't in a routing table you can see doesn't mean it isn't in a routing table or that it is wasteful. Continuing to pretend that it does will not make it so.

I would much rather see us use policy that leads to results such as how droughts have been handled in California (current policy text) rather than policy that leads to events like Tulip Mania and/or Enron. Proposal 151 represents a huge step in the direction towards Enron/Tulip Mania for the IPv4 address market.

>>> By allowing all transfers to be needs-free, we remove an artificial and un-needed market restraint whose existence will only serve to drive transfers off the books, with Whois accuracy diminished.
>> That market restraint is, IMHO, neither artificial nor unnecessary, but, rather very real and vital to preventing speculative hoarding of addresses placing the interests of greed over the efficient utilization of address resources. That is why I believe the entire premise of 151 is so completely flawed.
> Back to speculative  hoarding even though the hoarder is faced with an uncertain transition time, uncertain market conditions, uncertain governance, and a hard cap of controlling a maximum of .024% of total space. And of course none has ever been shown to exist, yet you risk Whois accuracy to maintain unnecessary roadblocks to natural transactions.

Hoarders are always faced with uncertainties such as the ones you describe. As to the hard cap, that's not a meaningful analysis of the situation. First, let's look at the total likely market of available addresses. The most generous estimate I've seen says that as many as 14 /8s might come available through market processes. That's only 6.3% of total address space to begin with. Of that, a /12 represents 1.7% of the total market, assuming the market reaches that most optimistic level of supply. Given that, the mere creation of 24 organizations would be more than enough to hoard enough resources to command 35%+ of the total market under proposed policy.

>>> I simply remind you of where this proposal started. It was seeking to answer the question "What would have happened in the MS/Nortel deal had not MS justified the transfer to ARIN?"
>> The transfer would not have been recognized in the ARIN database absent a court order to the contrary. No such court order has ever been granted and I have no reason to believe that one necessarily would have been granted in this case.
> Absolutely correct, ARIN would not have updated Whois, but Microsoft would own the address rights to blocks listed in Whois as belonging to ancient and defunct prior acquisitions of the defunct Nortel. This is not a desirable outcome, IMHO.

What rights? What are these alleged address rights? What do those rights mean?

You have no right to tell me that I can't use a number on my network no matter what the ARIN database or anyone else says.
You have no right to tell an ISP that you are not paying to route packets with those numbers in the destination field that they
must route those packets to you.
Address rights are, for the most part, a myth for all practical purposes.

What you have, instead, is a set of ISPs and others that generally cooperate based on a centralized registry of what the
community has chosen to recognize as authoritative information on who should be allowed to use particular numbers
for a particular purpose. Take away that agreement and the whole system devolves into chaos. The court cannot force
changes to that agreement, at least not easily, because it is generally out of scope for the court to make rulings forcing
entities that are not parties to a case to take actions in support of the case.

Let's suppose in the hypothetical situation that you describe, ARIN recognized that the previous address holders were
obviously defunct (through bankruptcy) and returned the addresses to the ARIN free pool. Since we're going hypothetical
here, instead of using a company name for the address usurper, let's call them Brand G.

ARIN now issue some of the addresses to Brand L, but, at the time informs Brand L that they might face issues with
Brand G as a result and explains the history. Brand L is sufficiently desperate for IP addresses that they choose to
accept the risk.

Brand L takes their whois registration and goes to their ISP and starts advertising these routes. ISP gets a call
from Brand G, whom ISP has no business relationship with. ISP can do one of two things:

1.	Tell Brand G. to take it up with ARIN, they're not registered in whois.
2.	Stop advertising the routes for Brand L.

My guess is that it's a bit of a toss up here and depends largely on the relative size and fear factors of
pissing off Brand G. vs. Brand L. as to which way said ISP is likely to go.

In the case where Brand G gets told to go away, a lawsuit likely ensues where Brand G names ARIN,
Brand L, and the ISP as defendants.

In the case where Brand L gets screwed, a lawsuit likely ensues where Brand G and the ISP get
named as defendants with Brand L as plaintiff.

What would happen in either lawsuit is virtually anyone's guess. I think that the probability of Brand G
getting injunctive relief protecting their rights in an unrecognized and unregistered transfer may
be pretty slim in most possible cases. In the case where the transfer was done by the bankruptcy
court, they might stand a somewhat better chance, but, even there, it's hard to see where ARIN,
the ISP, or Brand L would have any obligation not to engage in conflicting use of the addresses
under any existing statute.

That's the fundamental thing most people seem to miss about the internet. It all depends on everyone
cooperating about these things. Take out cooperation and the system rapidly devolves into chaos.
Nothing takes out cooperation like greed. Hence my desire to keep greed as much removed from
address policy as possible.

Frankly, I think that the market will likely destroy the IPv4 operating environment and lead to its
rapid disintegration. I think it may be one of the most traumatic and one of the biggest motivators
for IPv6 deployment. However, in spite of the fact that I think an unregulated market has great
potential to advance the interests of IPv6, I think the damage it would do is too great and should
be mitigated as much as possible.

Again, these thoughts are entirely my own and I alone am responsible for them. I am not a lawyer
and nothing above should be construed as legal advice or a legal opinion. It is just my observations
based on what I have seen and read.


More information about the ARIN-PPML mailing list