[ppml] Research on transfer markets, was: RFC 1744 and its discontents

[Tom Vest]

On Apr 18, 2008, at 11:26 AM, Scott Leibrand wrote:

> Tom Vest wrote:
>> On Apr 18, 2008, at 8:35 AM, Scott Leibrand wrote:
>
>>>
>>> One big overarching meta-assumption I think we might differ on is
>>> distinguishing what is ideal (what the men in suits could do with  
>>> the
>>> power of government behind them) and what is possible (what those  
>>> of us
>>> in jeans can do within our policy development framework).  For  
>>> example,
>>> if the government gets involved they can dictate how IP addresses  
>>> get
>>> privatized.  But as long as we think a non-regulated system is  
>>> better
>>> for the industry, we have to work within what's already been done  
>>> (for
>>> example, we can't force legacy holders to give back space, or charge
>>> high per-address fees to encourage conservation).
>> Hi Scott,
>> I think we all think of ourselves as realists/pragmatists, but that  
>> doesn't require us to be so fatalistic.
>> There's a vast middle ground between men with suits and guns and  
>> Hobbe's nasty, brutish (and short) competition of all against all.  
>> It's called "collective action" -- i.e., members of the community  
>> voluntarily, collectively deciding to undertake some action(s), or  
>> to forebear from taking some action(s), with the goal of achieving  
>> some outcome that is preferable to external occupation or complete  
>> anarchy. Wasn't that what happened when the community established  
>> the RIRs, and agreed to abide by the policies that community  
>> members themselves created?
>
> I agree that there is a middle way, which is why I'm so involved in  
> the transfer policy proposal.  I also agree that we can accomplish a  
> lot with collective action, and that we needn't involve the men with  
> suits.  I'm only pointing out that there are limits to what we can  
> accomplish with collective action, and that many parallels to prior  
> privatizations are of somewhat limited utility precisely because  
> this transition is occurring by collective action (without as much  
> government authority to force reluctant players to participate).

Hi Scott,

You may be right about the limitations of self-regulation/collective  
action in this case. However, those limitations do not alter in any  
way the "success conditions" that must be achieved, i.e., to avoid  
both the fire and the firemen. To expand on a comment from Yakov's  
March 1995 slides, we can choose to ignore the fundamental technical  
(and institutional, and economic) issues that will be affected by how  
we handle this, but that doesn't change the fundamental properties of  
those issues.


>>> If there was a land grab, it has occurred gradually over my  
>>> lifetime.
>> Can you tell us what "land grab" means precisely, and point us all  
>> to statistical evidence that that has already happened
>
> I don't think there has been a land grab: hence my use of "if".  I  
> think that IP addresses have, for the most part, been assigned  
> fairly and equitably based on the conditions and understanding at  
> the time they were given out.


I agree entirely. However, I also believe that resource transfer  
proposals as currently defined would represent a fundamental and  
irreversible break from this policy/practice/outcome.


>> and already happened to such a degree that everyone should dismiss  
>> the wholesale privatization of IP address resources as "nothing new"?
>
> You can call it a privatization if you want, but I think it's more  
> accurately just a distribution of IP addresses.


A "distribution" of resources that were previously defined and/or  
treated as common pool resources, which makes those resources  
alienable, and grants the initial recipients the right to choose the  
identity, timing, and terms of transfer to subsequent recipients is  
privatization, plain and simple.

There are cases where communities have willingly opted into similar  
arrangements, but I don't know of any where the contrast between the  
resource base (here, the entire final global supply of some critical,  
non-substitutable economic input) and the recipient community (a few  
thousand companies in business today, but not their stockholders, or  
their customers, or their customer's customers, or anyone else) is  
even remotely similar.


> What we're facing with IPv4 exhaustion is a requirement for  
> redistribution, which we can either meet with inaction (and possibly  
> end up with a Hobbesian free-for-all or men-in-suits regulation), or  
> we can collectively establish some standards for the redistribution  
> and continue the current cooperative self-governance that has  
> benefited the Internet so much for so long.

There are kernels of truth (or at least agreement) here, but many  
unanswered questions.
Whose demands for redistribution rise to the level of "requirements"  
-- aspiring sellers or would-be buyers? Incumbent buyers or new  
entrant buyers -- i.e., the current and future "customers" that don't  
participate in ARIN deliberations (yet)? Answering "all of the above"  
isn't really an answer, because the way the redistribution  
mechanism(s) are structured will largely determine how the interests  
and bargaining power of the parties will be defined and balanced, or  
not. Perhaps many people feel that questions of bargaining power  
shouldn't or can't be addressed by a collectively-defined transfer/ 
recirculation mechanism(s) -- i.e., that the market should be self- 
defining and self-managing. That's perfectly natural -- after all,  
incumbents always feel that way. But the Internet only emerged and  
grew in places where those wishes were not honored. Everywhere else,  
people had to settle for whatever Internet they could import (which  
benefited the other, market power-balancing economies tremendously).

How is any new entrant ever going to be able to bypass the need to  
reach the universe of IPv4-addressed resources? And if incentives for  
incumbent IPv4 resource holders are realigned so that the worse they  
ever face is big bucks for small IPv4 sales, plus the collective  
ability to manage/limit future competition, what incentive will they  
ever face to migrate to IPv6 themselves? Perhaps future new ISPs will  
take comfort from the old "donut hole" peering theories -- i.e., they  
won't really be totally marginalized forever -- but the fact is those  
theories only hold water in situations when/where (a) the supply  
critical resource (in this case, "private line" bandwidth inputs) is  
so vast and/or growing so fast that price discipline cannot hold, and/ 
or (b) incumbent resource holders are subject to gov-mandated resale  
requirements and price cap restrictions.

In the end I agree with you that a viable IPv4 recirculation mechanism  
would be a good idea. However, privatization without some novel,  
muscular enforcement mechanisms -- because the passive ones that have  
worked so well to date won't be enough -- is not going to work. It  
won't deliver the liquidity required (c.f., Michael Dillon's  
concerns), and if it does it'll either be because (a) prices will be  
high enough and beneficiaries skewed enough to draw an economic  
intervention or because (b) weak rules/enforcement  will drive many  
transfers "off the books", thereby steadily eroding the central  
registry (cue the national security intervention).

>>> In my mind, these are good and useful comparisons, but to my mind  
>>> they
>>> presuppose that the entity making the policy has authority to change
>>> allocations and/or privatize resources.
>> I'm curious -- are you suggesting that in the other cases, the  
>> policy entity had such "authority" but in this case the address  
>> resource community does not? To me, that seems like a strange (or  
>> maybe deeply depressing) position for an elected representative of  
>> the community to take...
>
> We have a collective action problem.  We have authority, but only so  
> far as the industry is collectively willing to cede it.  If we were  
> to assert authority to reclaim legacy space, or use fees as  
> governments use taxes, we would be exceeding our authority IMO.  If,  
> instead, we establish a transfer policy with reasonable restrictions  
> on behaviors that negatively affect the community, we are certainly  
> well within our authority.

I think that, within the limits of (external) law, the authority of  
the AC and other ARIN institutions is absolutely bounded by the  
interests and wishes of the community. However, those interests and  
wishes can and do change over time. I don't believe that the  
privatization of IPv4 would serve the long-term interests of the  
direct stakeholder community, or the much larger community of  
customers, future customers and address "needers", and Internet  
stakeholders at large. In the end I have to think that this won't be a  
minority position.

>> In any case I disagree with your claim. Different sectors and/or  
>> national economies can be looked at as "natural experiments."
>> You can look at the outcomes of individual cases -- where they  
>> perceived to be "successful" and by whom, who were the winners and  
>> losers, how did things look (x) years on, etc. -- all without  
>> getting stuck on how the initiatives got started. If you find that  
>> the results varied under different circumstances, then you can  
>> consider whether those results can be explained (i.e., dismissed)  
>> based on the the particulars of who initiated and drove the changes.
>
> I agree, so I guess I wasn't claiming what you thought I was.  :)

In that case my apologies ;-)
I will work up some cases studies of past big/important initiatives  
that are are quite comparable, e.g., ones involving simultaneous  
initiatives to alter the status/treatment of former common pool  
resources, plus the transformation of mechanisms and terms of exchange  
involved in the distribution of those resources. Will share the  
results here as soon as possible...


>>> Perhaps another angle would be to look at the dynamics of real-world
>>> transfer markets.  I think the closest analogy would be to real  
>>> estate.
>>> For starters, our situation is much like that of a number of
>>> developing countries where residents didn't have title to their  
>>> land.
>>> In some cases, when they were given title, and a working system, and
>>> were able to do all the good things you'd expect (buy & sell land,  
>>> use
>>> it as collateral for loans, etc.).  In other cases, the system  
>>> didn't
>>> work well, transactions were not recorded properly in the central
>>> registry, no one knew exactly who had what, etc.  I'm not sure what
>>> exactly the differences were between the systems that worked and  
>>> those
>>> that didn't, but I'm pretty sure a researcher could tease out the
>>> distinctions and come up with some lessons for how to structure a
>>> transfer system for IP addresses.
>> Land is a great comparison -- but not in modern times, when  
>> ownership and pricing is transparent, and there are lots of jobs  
>> other than being a farmer, and political and economic self- 
>> determination are not inextricably linked to land ownership. The  
>> right comparison is land, and the role of land and land markets in  
>> the 19th century and earlier. That's when land was "non- 
>> substitutable", just as IPv4 is today.
>
> And I would certainly love to see the results of some analysis of  
> the redistribution of land from feudal lords to private landholders.

Basic story: It only happened after land ceased being a non- 
substitutable bottleneck input for economic and political life (e.g.,  
as semi-independent towns and non-agricultural sources of capital  
formation emerged). It took a very long time in places where it didn't  
happen through violent/revolutionary redistribution. In a lot of  
places where land and agriculture continued to be the primary economic  
sector even after the feudal era, de jure land rights don't matter  
much, as a small number of incumbent land holders were able to  
leverage new market mechanisms to (re)consolidate their positions --  
maybe not forever, just for all modern history to date.

> However, I don't think that the modern situation is any less  
> relevant. You mentioned transparent ownership and pricing: I think  
> both of those are achievable for IPv4 addresses.

Today, transparent ownership is (more or less) revealed by whois,  
which continues to be (more or less) complete and accurate for those  
LIRs that send annual membership keep-alives, which they send because  
they expect to need subsequent allocations, which makes them expect to  
be beholden to future RIR policies, which keeps them (more or less)  
engaged in the RIR policy development process; repeat as necesaary.  
Much of this is not true for legacy, RSA non-signatories, and I  
believe that accuracy/completeness of whois for them is substantially  
lower than the for RSA signatories (ARIN staff can confirm/deny this  
if there is any doubt).

The point is, none this will continue to be be true for anyone if IPv4  
is privatized.

Today, transparent pricing is achieved by making everyone subject to  
the same categorical eligibility and pricing rules, and by prohibiting  
third-party resource transfers. How will pricing transparency be  
achieved in a decentralized IPv4 market -- esp. in an industry where  
prices for every other kind of input and transaction are closely  
guarded by NDA?

Many people have indicated that a de facto black market exists today  
-- but no one has provided much of an indication of prices available  
in that market, much less comparative/average prices.* Sure, this is  
due in part to the illicit nature of the transfers today -- but no one  
I talk to in the ops community ever puts much stock in price surveys  
for open market inputs (bandwidth, colo) -- even the ones that some  
people pay thousands of dollars for. Is that the kind of (nudge nudge  
wink wink) transparency that can be expected?

*The numbers I've heard make me believe that black market prices now  
are much higher than RIR fees, reflecting the value of "non- 
encumbered" IPv4 that is not subject to RIR/community policies and  
requirements -- even though such resources have no official  
provenance. If that's correct, it doesn't bode well for matters in a  
post-privatized world.

> As far as substitutability, I think we do have a viable substitute:  
> IPv6.  It's not perfect, of course, but substitutes rarely are.

It's substitutable enough for incumbent IPv4 holders. It's not  
substitutable for anyone else. It never will be.
We believe that this won't be a fatal condition because we assume that  
enough IPv4 will remain accessible on reasonable terms so that new  
IPv6 native operators will be able to control their own v6/v4  
translation gateways. Those gateways will be an eternal requirement  
for all future operators unless/until the universe of IPv4-numbered  
users, content, etc. is renumbered into IPv6 -- or they become  
completely irrelevant (c.f., the donut theory). Now, why would  
incumbent IPv4 dealers (and/or aspiring speculators) ever want either  
of these things to happen?


>>> I don't claim to know how to do (or direct) research, but I
>>> do know that we need to focus our attention on the things that are
>>> relevant to what is possible under the RIRs policy development  
>>> processes
>>> (for now),
>> This is potentially very useful advice. However, to operatioanlize  
>> it, please clarify:
>
> I can't fully specify the possible or the impossible, but I can give  
> some illustrative examples.
>
>> 1. what is possible under the RIR policy development process?
>
> - Any of the spectrum of transfer policies, including APNIC's wide- 
> open one, RIPE's LIR-only one, or ARIN's more restrictive version.
> - Doing nothing, at least for awhile.
> - Some attempt at "soft landing", including policy forcing  
> recipients to start transitioning to IPv6 to get more IPv4 space  
> after a certain point.
> - etc...


I wonder: if the community ultimately concludes that none of these  
things will be sufficient to prevent the oft-mentioned unhappy  
outcomes, will policy development alternatives still be restricted to  
these narrow options?


>> 2. what is impossible under the RIR policy development process?
> Or, more particularly, what policies would be impossible for ARIN to  
> enforce?
>
> - Forcibly reclaiming addresses from legacy holders, or imposing  
> substantial fees on them.
> - Imposing substantial fees on large IPv4 address holders (or  
> substantial per-address fees across the board) in an attempt to  
> encourage conservation.
> - Repurchasing addresses from current address holders to replenish  
> the IPv4 free pool.  (Perhaps not completely impossible, but  
> extremely problematic.)
> - etc...


I wonder: if the community ultimately concludes that some of these or  
other similarly challenging commitments, may be necessary to forestall  
TEOTWAWKI, what prevents them from exploring policy development  
alternatives like these?


>> Or to be more pointed:
>> 3. Is collective action still possible under the RIR policy  
>> development process?
>
> Yes, with limitations on what can be achieved by collective action,  
> as described above.


I think it's time for every member of this community to start  
calculating the enterprise-level net present value of preserving  
industry self-governance, as well as the long-run cost of losing it --  
and then publicly indicating what kinds of policy initiatives they  
think are "worth" considering.


>> Once that is clarified, figuring out what is/is not relevant should  
>> be easy enough...
>>> and particularly on research whose outcomes include
>>> recommendations for how we can improve our policies to help steer a
>>> middle course between potential negative outcomes on both sides.
>> That's definitely a goal that we share.
>
> Yeah, it's nice that at least we all seem to be working for many of  
> the same goals.
>
> -Scott