February 2nd, 2009 by Gordon Cook
Author’s Note: This essay appears on pages 15-18 of the March 2009 COOK Report on Internet Protocol.
A Tipping Point for the Internet?
Catching the precise moment of a tectonic shift in a global system as large and important as the Internet may be viewed as an exercise in the improbable. However, I point out in this summary that I think we are precisely in the midst of such a shift. The largest portion of this March 2009 COOK Report issue is approximately 20,000 words of discussion of the ramifications of the exhaustion of the remaining pool of routable IPv4 address blocks - (pp. 30-63). This article serves as an introduction to that detailed discussion.
The RIPE policy announcement of December 16, 2008 sets conditions by which RIPE members who have IPv4 address block assignments can reassign some or all of them to other members. Although the action, at first glance may seem trivial, it is happening only because the pool of assignable IPv4 numbers, a limited resource on which internet growth depends, is running out. As the discussion in this issue points out, this change in policy has ramifications of which almost all of those who depend on use of the internet are unaware. This shift in the permissible use of the Internet’s most basic economic resource will have profound consequences. It is tantamount to the adjustment of stresses deep with in the San Andreas fault. The new policies will begin to send tremors through the global system in ways that bode ill for the open and competitive internet we have known so far.
IPv4 numbers are the fundamental building blocs of the global internet. While people can “participate” in the direct provision of internet content offering and web hosting based on transitory assignments of IPv4 numbers, the assignment of IP numbers on a permanent basis from a Regional internet Registry (RIR) is the only way a business can enable itself to route its customer’s traffic via an ASN number. Such capabilities establish such a business as one of about 30,000 independent providers of Internet service in the world.
With the beginning of the Internet in the 1980s, a handful of universities and large corporations were able to participate in the deployment of TCP/IP as a transparent overlay of carrier networks. The protocol was hardware independent in a way that that other networking protocols were not. Because of this independence, the university community and its technology partners were able to construct an inter-network of networks on a large and rapidly growing scale. US government policy under the leadership of the NSF enabled the NSFnet backbone to interconnect to foreign networks in the late 80s and early 90s. At the same time, with a loosening of acceptable use policy, the NSF enabled small scale dial up commercial providers to connect to university endpoints. The university based Internet began to morph into the commercial Internet.
With the advent of the commercial Internet as marked by the decommissioning of the NSFNet backbone on April 1, 1995, the number of service providers, by then in the hundreds, quickly increased to thousands of independent ISP businesses. To facilitate the process of IP number assignments (the street addresses for the delivery of ‘packets’), regional internet registries (RIR)s were established. In turn these registries were used by their members to administer the policies that they, the members, established for the gradual assignment of the IPv4 blocks of numbers needed by newly formed ISP business. These were businesses that wanted to become independent economic participants in the rapidly growing protocol overlay being constructed with a diverse mix of technology inputs that included campus and corporate LANs, new green-field point-to-point facilities, and fractional infrastructure products bought or leased form incumbent carriers.
The IPv4 numbers were simply an indispensable part of the TCP/IP protocol. To set up a service using TCP/IP, the service providers needed to be able to assign their customers unique IPv4 numbers from which and to which packets could be sent. The use of the IPv4 address blocks of specific sizes was supported on the basis that the ISP needed that many IP addresses — no more, no less — in order to connect useful things to the Internet, and that particular allocation was justified for as long as that need remained — no more, no less. If the business was disbanded, the IPv4 blocs had to be returned to the registry to be reassigned. The blocks were not owned, They were not property. In economic terms, they were in effect “inalienable” — not subject to being sold or transferred by the original allocation recipient to a third party, or to being purchased or acquired in other ways by a third party.
The blocks were like spectrum frequency assignments before frequency was auctioned. While spectrum frequency was there to be used according to the rules of the regulator, the IPv4 blocs were there to be used according to the rules set by the ISP members of the RIR. In the case of radio you broadcast on the frequency. Before auctions you could not claim to own the frequency. In the case of the ISP, IPv4 numbers are obtained directly from the reserve address resource pool administered on behalf of the ISP community by the Regional internet Registry, or alternately from a RIR member ISP. This hierarchical arrangement, with neutral RIRs at the top, competing ISPs at the next level, and individual users below that, parallels the organization of the banking sector, with a central bank (or occasionally, a “banker’s club”) at the top, competing lending institutions at the next level, and aspiring borrowers below, according to Tom Vest.
Vest suggests that this symmetry is no coincidence, but rather a product of the fact that IP addresses performs the same kind of “medium of exchange” function that money plays in the conventional economy — and that the uses of IP addresses are subject to the same kind of systemic risks that can render money useless in certain circumstances in the wider world, for example, in times of extreme inflation or deflation.
Consequently IPv4 number assignments had economic utility by virtue of the fact that they were an indispensable part of the TCP/IP protocol that could be used as a transparent overlay” technology across carrier networks. The overlay was transparent to the networks that did not distinguish data network protocols from voice. Inter networks could be built atop various (telco) inputs, generally without requesting or securing their explicit permission about how the inputs would be used. Why? Because TCP/IP was just one of many data protocols to which the telcos were obligated to provide common carriage.
For the 1980s all this worked well. Nevertheless, as long as IP addresses are essential but scarce, the prospect of a single entity having control of IP addresses created an inescapable conflict of interest for competing service providers. Recognizing this fact, beginning in 1993, ISP communities started establishing quasi-independent central bank-like institutions — “regional internet registries” or RIRs– to administer the distribution of these critical resources for the purpose of connecting useful things to the Internet.
The Tectonic Shift
Until December 2008, IP addresses distributed via the RIR system could not be bought or sold; i.e., they lacked the quality of “alienability,” which economists regard as an essential feature of private property. But this lack was itself a feature – not a bug. IPv4 could be used only by an entity that agreed to create value by actively contributing to the Internet system. IPv4’s lack of any other kind of use value, and the conscious collective decision to prevent it from acquiring exchange value, were the primary causes and rationale for creation of the RIRs and the “needs-based” allocation regime in the first place.
However, IP addresses can also have an “exchange value” if they have become scarce; that is, if those that actually need IPv4 now to attach useful things to the Internet can only obtain addresses from someone else that needs them less, and thus may be persuaded to part with them for some consideration. This is the strategy that was, in effect, chosen by the RIPE community when an IPv4 resource transfer policy was approved on December 16, 2008.
It was not the only conceivable strategy. For example, if the successor IPv6 addressing format had been transparently adopted by most or all community members, the scarcity and intrinsic exchange value of IPv4 addressing would have disappeared. But that did not happen.
Moreover, the fact that IPv6 has been rejected has even more far-reaching implications. Making IP number resources “alienable” strips them of one of the critical features that previously made them (1) irrelevant and transparent to countries and national jurisdictions, and (2) effectively manageable through voluntary, “self-governance” mechanisms.
If and when IP numbers become alienable, economic substance doctrine will eventually dictate that they ARE property for all practical (and legal, and regulatory, and taxation) purposes, regardless of whether some people might want to claim otherwise. The reason for this conclusion is that, given the economic impact of these markets, there will be litigation. When litigation occurs that the transaction (address block transfer) must have a meaningful economic purpose to be legitimate or sustainable in a court of law. An affirmative answer means that the IP Block transfer is one that involves property (something of economic value to the possessor. Going down this road invites government involvement. Because when IPv4 assignments become property, the only entity that can tell you what to do or what not to do with your private property is the property rights guarantor, i.e., the government.
Given the new opportunity to “own” IP addresses, how are incumbent services providers — and especially incumbent facilities-based carriers — likely to respond? The outcome of another recent and relevant privatization initiative — i.e., U.S. spectrum auctions — suggests a likely, if chilling, scenario. Given the benefits of securing scarce resources for their own customers, but even more importantly of blocking any possibility of competitive bypass, one may assume that incumbent territorial facilities owners will do everything they can to acquire all available IPv4 addresses. Success would empower them to become the unilateral arbiters of all uses of the TCP/IP overlay — they will be able to demand whatever share they want of any IP-based service that they permit, and absolutely preclude any services that they dislike.
Even if this scenario seems too pessimistic, it is hard to avoid the conclusion that, after all the usable IPv4 blocks have been distributed (circa 2011), the inheritors of RIR-era IPv4 will literally possess the keys to the kingdom. They will stand to achieve and enjoy permanent market power simply by doing nothing. That could mean that the “open Internet” could be finished forever — or at least that the next moment of openness may come only after some new technology is invented that makes it possible to bypass TCP/IP in the same way that the latter made it possible to bypass the arbitrary restrictions imposed by telco facility owners.
The prospects currently look grim. On 24 January 2009 the ARIN Advisory Council (AC), acting under the provisions of the ARIN Policy Development Process, recommended that the ARIN Board of Trustees adopt: Draft Policy 2008-6: Emergency Transfer Policy for IPv4 Addresses
This policy document will put in play a process even less transparent than that created by the members of RIPE. The only apparent way out might be for the registries and IANA to step back from the brink and establish procedures by which new entrants could be very gradually allocated the remaining numbers.
Where Do We Go Now?
No, the Internet will not disappear, it will very likely however become much more expensive to use. It will also likely fragment and lose much of its ability to stimulate growth and innovation. Once a property right is recognized, history suggests that its beneficiaries rarely give it up willingly. If IPv4 addresses become property, then there will be no IPv6 transition, or any other successor addressing format or technology that can succeed without the active support of future IPv4 owners — or at least none that might undermine the advantages and rewards of IPv4 ownership. To expect otherwise would be to assume a level of altruism that has been scoffed at as unrealistic by the current generation of IPv6 refuseniks. If a transition is contrary to their individual private interests today, it will be doubly so if and when the only globally interoperable IP addressing resources are their exclusive property.
In this new uncharted territory, anticipating the most likely course of Internet development is incredibly daunting. And yet, with the stakes as high as they are it would be reckless to simply trust to fate that we will get to the other side with all of the Internet’s critical features intact, if not improved. A hasty, incautious reading of the new RIPE and ARIN policy developments might lead one to assume that that nothing much of significance has changed. However, once one understands the dependencies between IP address allocation, registration, and address uniqueness (without which the Internet stops working) — and the links between neutral, eligibility-based allocation practices and overall industry openness, this seemingly modest little change takes on the greatest, broadest possible significance.
Of course, the new “resource transfer” policies won’t necessarily lead to industry closure, or to the inevitable erosion of the presumed uniqueness of individual “public” IP addresses — but it’s not clear that implications of these possibly existential risks have been given all due consideration. If managed properly, perhaps a better, more open — possibly IPv6-based — Internet will result. If not, the scarcity of globally routable IP addresses could easily become the greatest bottleneck to continued growth and evolution in the Internet’s brief history.
The choices that will determine which course we all will follow are being made right now. Everyone who has any interest in the Internet, today or in the future, should take note, and if necessary speak up — now, while the future still remains malleable.
Acknowledgment: Thanks to Tom Vest who gave valuable assistance in getting all this in focus, and clarified many historical and technical details. Editorial comments or interpretations, especially regarding specific RIRs or RIR policies, are my own.