[arin-ppml] Fwd: ARIN-prop-165 Eliminate Needs-Based Justification
astrodog at gmail.com
Fri Feb 24 13:56:09 EST 2012
On Fri, Feb 24, 2012 at 12:45 PM, Tom Vest <tvest at eyeconomics.com> wrote:
> On Feb 24, 2012, at 4:00 AM, Astrodog wrote:
>> On Thu, Feb 23, 2012 at 3:59 PM, John Curran <jcurran at arin.net> wrote:
>>> Milton -
>>> Is there an fairly straightforward economic explanation that addresses
>>> the questions that John Sweeting raised? If so, could you outline it?
>> Sorry for hijacking a bit, but I do have an answer.
>>> From: "Sweeting, John" <john.sweeting at twcable.com>
>>> Subject: Re: [arin-ppml] ARIN-prop-165 Eliminate Needs-Based Justification
>>> Date: February 17, 2012 12:55:52 PM EST
>>> To: Milton L Mueller <mueller at syr.edu>, 'Paul Vixie' <paul at redbarn.org>,
>>> "jeffmehlenbacher at ipv4marketgroup.com"
>>> <jeffmehlenbacher at ipv4marketgroup.com>
>>> Cc: "arin-ppml at arin.net" <arin-ppml at arin.net>
>>> Hi Milton,
>>> A few questions since I am not an expert in markets. Do you think there is
>>> a scenario where the largest and wealthiest of the world's ISP's would buy
>>> up as much address space as they could just to keep it out of the hands of
>>> their competitors? Or new start ups in their markets? There are some very
>>> wealthy companies out there that would not blink an eye at spending the
>>> money needed to run their competitors out of the market but maybe that is
>>> not possible in this particular market. Is there a way that a small but well
>>> of portion of the market could insure prices are so high that the number of
>>> competitors would be capped or even reduced? As I said I am not an expert on
>>> markets so would very much welcome your views on this.
>> It's very unlikely that large ISPs would make this sort of move for a
>> number of reasons:
>> First, it should be noted that even the largest ISPs are not
>> particularly well positioned financially for this sort of endeavor.
>> AT&T as an example, has ~$3.2 billion in cash, and after paying
>> dividends, has lost money somewhat the past few years. When compared
>> to some of the end users of address space, they simply wouldn't be
>> competitive bidders, limiting how much of the "for sale" capacity they
>> could acquire. Aside from this, the largest ISPs cannot cooperate on
>> their purchases and would be at a significant disadvantage compared to
>> their large competitors who did not attempt to drive up the price.
>> It is also notable that large ISPs are not precluded from doing this
>> under current policy. Simply through changing their address scheme for
>> CPE, most large ISPs could nearly double their current allocation.
>> Considering that in 8.3 transfers, they only need to show 24 month
>> need, most could go significantly past this (for example, planning to
>> offer customers public IPs for all of their various bits of network
>> enabled equipment... note that under current policy, they do not have
>> to actually do this. Merely intend to.) As a result of how rare
>> auditing is, its unlikely that they would see revocations, either.
>> Under current policy, a small ISP attempting to become a large one
>> would face a much greater regulatory hurdle.
>> I think there is consensus around the fact that long term, IPv4
>> allocations are likely to have little to no value. As a result, they
>> are particularly unattractive as speculative investments. Someone
>> holding a large number of addresses without utilizing them as an
>> investment is taking a very large risk and would be very unlikely to
>> hold those addresses for long, due to their value trending to zero.
>> The same reasons why many of you have expressed... acceptance with
>> removing needs requirements in a few years would make speculation on
>> IP addresses today minimal.
>> Finally, a major way to keep a handle on speculation would be for ARIN
>> to hold some allocations back, post-exhaustion, with authorization to
>> release some of those addresses if it appears that an entity is
>> attempting to "corner the market" at a given point in time. Another
>> option would be verbiage specifically prohibiting speculation, through
>> post-purchase auditing. These do not need to occur frequently to
>> eliminate speculative behavior.
>> --- Harrison
> Hi Harrison,
> That's a fine set of arguments, and I do hope that you are right, but it might be worth considering the following:
> 1. Your examples don't capture the most compelling incentive(s) that might hypothetically motivate an operator's attempt to strategically leverage IPv4 scarcity for commercial/competitive purposes. The Internet has a strong tendency to "route around" congestion, censorship, scarcity, and other bad things, not only (or primarily) because the technology just works that way, but because clever operators are constantly, actively looking for opportunities to optimize their network cost/performance characteristics -- i.e., by topologically "bypassing" such problems. Of course, network input suppliers (the majority of which are now also "operators" themselves) are fully aware of this fact, and as a class they naturally tend to be reluctant to compete on price ( -> commodification) *if* there is any other alternative. Consequently, smart input suppliers are constantly trying to identify and "capture" profitable market niches that are very likely to remain profitable because operators cannot easily bypass them -- i.e., because they constitute a "bottleneck" to profitably providing network service to some place/population. Sometimes (increasingly uncommon IMO) such bottlenecks are be embodied in some unique technology or non-fixed asset. Today, in this industry, such bottlenecks are far more likely to be explicitly or implicitly spatial/geographic in nature, e.g., a particularly attractive room in a building (e.g., an IXP/IDC), building in a particular neighborhood, neighborhood right-of-way in a city, city in a country, etc., or indirect access to the same via interconnection/traffic exchange with another operator that has better or perhaps exclusive topological access to to the desired "turf."
> 2. At various times/places, IPv4 can and has embodied an absolute bottleneck of the former (non-fixed) kind, at least for some network operators and/or would-be operators. The possibility (or rather "eventuality") that IPv4 would become a bottleneck was one of the factors that drove the creation of the RIR (needs-based) allocation system, and the limited accessibility of IPv4 on non-adversarial terms to aspiring network operators outside the US was one of the demand drivers that led to the spread of that address distribution model to every continent. Today, thanks to NAT, RFC1918, and IPv6, IPv4's status as a "quantitative bottleneck" to delivering many kinds of services has diminished substantially -- i.e., life might be vastly easier if one had more IPv4, but -- assuming that you have any at all -- an independent network operator can generally remain in business as such, and even continue growing, by strategically reassigning/multiplexing and/or supplementing their limited IPv4 holdings with private addressing and/or IPv6. However, for the overwhelming majority of places and aspiring (future) service providers, IPv4 still remains an absolute/qualitative bottleneck to entering (appx. any/all) Internet services market, at least with the same sort of "independent" status that all of one's incumbent competitors would enjoy.
> One day that might not be true, but when (or whether) that day might come is anyone's guess. You may be right that there is "consensus around the fact that long term, IPv4 allocations are likely to have little to no value," but if so that consensus is overwhelmingly among individuals that work for institutions that (a) have to-date put little if anything at risk to hasten that outcome, and (b) simply by virtue of their possession of IPv4 could have quite a lot to gain if the consensus prediction turns out to be wrong. Obviously there are some exceptions, e.g., networks that have literally bet the farm on IPv6, but they represent a tiny minority among those holding the "consensus" view.
> Note that I'm not suggesting that people are misrepresenting their actual beliefs, but it's a lot easier to have strong opinions about future unknowns when being wrong is a (privately) cost-free option. Like I said, I do hope that the consensus is right, but caveat emptor is always good advice.
> 3. Given the fact that, by itself, a strategy of "IPv4 hoarding" seems unlikely pay off in the near future, and might never become an effective means of *establishing* market power (i.e., except under relatively improbable circumstances/scenarios, for the perennially modest market segment of aspiring new entrants), but arguably has already/repeatedly proven to be quite effective as a strategy for *enhancing* market power under certain conditions (e.g., during the preRIR/early RIR era, in regions/bargaining situations where BYO addressing is unacceptable, etc.), one should assume that any shift toward strategic "IPv4 hoarding" would probably start -- and if left unchecked, end -- with the kind of entities that already control some other critical bottleneck(s) to Internet service delivery, as described in  above.
Thanks for the explanation.
Those are definitely significant concerns, but I'm not sure how needs
testing addresses them. Large ISPs already have the ability to horde
address space, in that they can make relatively minor changes that
would suddenly make them "need" (or, at least, qualify for) quite a
bit more address space than they already have. These entities are also
the ones most likely to be well versed in how to meet the letter of
ARIN's requirements. It seems like the needs testing simply adds one
more advantage they would have over their smaller counterparts and
Perhaps the solution to this problem is a reserved block for small
allocations, where the "reservation" survives 8.3 transfer?
At least in a more open market, such entities may also decide that
they can make more money using their space more effectively, and
selling the extra on, post-exhaustion, though I'll admit this seems
like a fairly unlikely outcome.
More information about the ARIN-PPML