[arin-ppml] [address-policy-wg] The price of address space

tvest at eyeconomics.com tvest at eyeconomics.com
Thu Jul 23 08:14:41 EDT 2009

On Jul 22, 2009, at 12:32 PM, Jim Reid wrote:

> On 22 Jul 2009, at 16:18, <michael.dillon at bt.com> <michael.dillon at bt.com 
> > wrote:
>> Where is the industry going to find 1.2 billion dollars to sustain
>> growth of the network after IPv4 runout? And where is a new entrant
>> going to find $100,000 to buy their first allocation, assuming that
>> the price doesn't rise even higher when the free alternative no
>> longer exists?
> I'm reminded of the late 90's .com joke/business model about how to  
> becomea multi-billionaire: Set up a company with 100Bn shares. Sell  
> one share to yourself for $5. Congratulations! Your company is now  
> worth $500Bn. So exchange your company's paper for a medium-sized  
> country.
> To be serious, I think it's unwise to extrapolate from your anecdote  
> about someone offering a six figure sum for a /20. Or to use that as  
> the basis for a valuation of the allocated and/or yet-to-be- 
> allocated IPv4 address space. For one thing, who could possibly know  
> for sure if the figure you quoted that was offered recently will be  
> the going rate for a /20 after the IPv4 run-out? If someone can do  
> that, please provide me with the winning numbers for next week's  
> lottery.... :-)
> If we assume there will be a market in IPv4 after the run-out, we  
> should expect the usual laws of supply and demand will mean there's  
> some sort of price equilibrium in the market.

Hi Jim,

I'd have to say that that depends a lot on what you mean by  
In principle, in an IPv4 transfer market that has some aspiring buyers  
and some aspiring sellers sellers, some transfer transactions may  
occur, and any average of the sum of prices paid that you find  
meaningful (e.g., price per /32, or any other prefix size of interest)  
could be described as the equilibrium price at that point in time.

In practice, however, that's a entirely "academic," if not utterly  
meaningless statement. In the absence of a mandatory central clearing  
house for all transfer transactions, the only "information" on  
prevailing prices that will be available to anyone -- buyers or  
sellers -- will be whatever whispered rumors you can pick up in the  
halls of the next ops or IETF or policy meeting. Of course all  
transfer transactions will inevitably be covered under NDA, and so the  
only parties who might hypothetically have an incentive to share  
information (i.e., shell-shocked buyers) will be officially unable to  
do so. So there will be an equilibrium price, but it will be about as  
accessible to you and I -- or to any aspiring buyer or seller -- as  
the last digit of pi (or about as knowable as it is today, in the  
shadow market that some have alleged already exists).

But even setting aside the problem of pervasive information  
asymmetries, there are several reasons to assume that the likely  
equilibrium price might be higher than you expect.

First of all, know that at almost every point in time (until something  
very dramatic changes), most if not all potential IPv4 sellers are  
very likely to maintain a steep "reservation price" for IPv4; those  
that don't absolutely require a high minimum will be cleaned out as  
quickly as they enter the market -- by buyers who will act this way,  
if they're willing to sell at all. Second, any sign of substantial  
ongoing demand for IPv4 transfers among established, IPv4-based  
operators is likely to have a perverse effect on the minimum  
reservation prices of any remaining potential IPv4 sellers. Consider:  
today some operators express confidence in an imminent future of  
relatively transparent IPv6-based inter-domain routing -- i.e., one in  
which they will not suffer commercially if they can only offer IPv6 to  
their new and still-growing customers. That confidence is  
(necessarily) founded on the assumption that most if not all other  
operators will either be making their customers and resources  
accessible via IPv6 (one way or another), or else will be exiting the  
market. What will happen to that confidence if/when those IPv6- 
oriented pioneers observe that some of their peers are willing to pay  
a high price for any surplus IPv4 that they'd be willing to sell? How  
many of then will assume that the aspiring buyers are all suckers who  
are doomed to failure because of their apparent unwilling to  
incorporate IPv6 promptly? Alternately, how many will begin to wonder  
whether they should raise their reservation prices even higher --  
either because the demand is there, or because their certainty in that  
IPv6-based future is marginally undermined? How many would decide to  
withhold their salable IPv4 altogether, at least until the future  
becomes a little less cloudy...?

And of course, if there are even a small number of speculators in the  
market, these doubts are likely to be greatly amplified. A speculator  
doesn't care about which addressing format is better for customers,  
because they don't have any. For a speculator, the relative merits of  
IPv6 actually represent a threat, because the more compelling those  
merits are, the less profitable / less durable their IPv4 brokerage  
business is likely to be. If enough speculators are able to  
participate in the market, their preferences could have a big impact  
on which addressing standard(s) -- if any -- will be operationally or  
commercially viable in the future. And there is no barrier to deter  
speculators from entering the market, other than the willingness of  
service-operating IPv4 buyers and sellers to altruistically refuse to  
deal with them (i.e., even when they make the best offers).

Granted, these observations reflect the kind of situation we're facing  
today, when very few resources are attached to the Internet with any  
form of IPv6, and very very few (if any) of those that are IPv6- 
attached are reachable from any other routing domain via any means  
that is not absolutely dependent on IPv4. People often point to this  
caveat by way of implying how different it will be when IPv6 is more  
widely deployed -- seemingly without noticing that such statements are  
tautological. Things will be different when they are different, no  
doubt -- but how will we get from here to there? Who's going to lead  
that charge so that others can make such statements with semantic  
content > 0 ?

I have heard quite a few current operators -- by definition, all IPv4  
holders -- suggest that post-IPv4 allocation-era new entrants will be  
the primary drivers for this transition. But from what we know from  
the allocation records and the routing table, etc., a nontrivial share  
of people saying (thinking?) this today must be among the vast  
majority of operators that do not currently offering native IPv6  
routing services. Perhaps there is no contradiction, and people are  
saying/thinking such things because, today, they fully expect to offer  
some form of IPv6 as soon as that must-have-IPv6 demand grows.  
However, when that time does arrive, will prospective upstream  
providers have an incentive to offer native IPv6 routing -- i.e., the  
kind that would provide the IPv6-only newbie with the same kind of  
status that they themselves enjoy (where the lines between customer  
vs. peer vs. provider status are dictated by  size and contract, not  
by inheritance or privileged access to a closed, proprietary  
technology)? Or, will the incumbent incentives favor offering the kind  
of IPv6 routing service that makes it substantially more difficult, if  
not structurally impossible, for the new entrant to organically  
transition into inter-domain peer or routing service provider roles as  
growth and other circumstances permit today?

If private commercial incentives favor the latter in many/most cases,  
is competition from other, more IPv6-friendly routing services  
providers likely to be sufficient to ultimately tip the balance in  
favor of substitutability of IPv4 and IPv6 in inter-domain routing?  
It's anyone's guess, but I think some insight might be gleaned by  
talking to, for example, tier-two and smaller providers in Latin  
America and Africa that were operating before the establishment of  
LACNIC and AfriNIC -- i.e., in a times/places when new entrants had  
little or no choice but to obtain address resources as well as routing  
services from an incumbent commercial operator.

> The point I'm making is that it doesn't seem sensible or even  
> possible to predict what these prices might be post run-out. Or  
> worry too much about that. Or to invent policies which somehow  
> influence or control those prices. [That might well have the look  
> and feel of a cartel.] It would be an interesting academic exercise  
> for economists to model the impact of various pricing scenarios  
> though I'm not sure how useful that would be in practice. That said,  
> it would be nice to have some sort of idea of the price points where  
> the trade-offs between buying IPv4 or using more NAT/ALG or  
> deploying IPv6 start to get fuzzy.

I agree wholeheartedly that quantitative models that seek to predict  
point-specific IPv4 transfers prices are not constructive -- if only  
because no one who really could use such information is likely to  
trust them, no matter how accurate they might actually be. That said,  
I think that the application of a little common sense plus industry  
experience/know-how to the basic question can take one a long way  
toward understanding what kinds of outcomes are marginally more vs.  
less likely, and perhaps which are quite unlikely.

> Any market that develops should be self-correcting.

That is axiomatically true, but then there are some self-corrections  
that may not be appealing to all stakeholders.

For those who are indifferent to the question of whether or not the  
inevitable correction will result in a future in which the IPv4/IPv6  
distinction disappears, and how (e.g., because IPv6 becomes a viable  
substitute for IPv4, or because IPv4 lock-in makes it a permanent  
mandatory requirement for any aspiring routing services provider),  
there is a lot less to worry about.

Others might want to give it some thought.


More information about the ARIN-PPML mailing list