________________________________________________________________________ prop-050-v004: IPv4 address transfers ________________________________________________________________________ Author: Geoff Huston gih@apnic.net Version: 4 Date: 6 March 2009 This proposal contains the transfer proposal features that reached consensus following community discussion at the APNIC 27 Policy SIG on Thursday 26 February 2009. 1. Introduction ---------------- This is a proposal to amend APNIC policy restrictions on the transfer of registration of IPv4 address allocations and IPv4 portable address assignments between current APNIC account holders. This proposal is a refinement of the historical resource transfer policy and applies to IPv4 resources held by current APNIC account holders. 2. Summary of current problem ------------------------------ Current APNIC policies relating to the registration of transfer of address holdings limit the eligibility of registration of transfers to those relating to mergers and acquisitions of entities that are administering an operational network. It is currently anticipated that the IPv4 unallocated address pool will be exhausted in a timeframe of between 2009 and 2011. There is a very considerable level of investment in IPv4-based services in the Asia Pacific region, and a transition to IPv6-based service delivery is likely to take longer than the remaining period of unallocated address availability. Accordingly, it is likely that demand for IPv4 addresses will continue beyond the time of unallocated address pool exhaustion, leading to a period of movement of IPv4 address blocks between address holders to meet such continuing demand for IPv4 address blocks. It is the objective of the APNIC IPv4 address registry to accurately record current address distribution information. This proposal's central aim is to ensure the continuing utility and value of the APNIC address registry by allowing the registry to record transactions where IPv4 addresses are transfered between APNIC account holders. 3. Situation in other RIRs ---------------------------- RIPE has implemented a transfer policy in its region. See section 5.5, "Transfers of Allocations", of: IPv4 Address Allocation and Assignment Policies for the RIPE NCC Service Region http://www.ripe.net/ripe/docs/ripe-449.html#55 ARIN's Advisory Council has recommended that the ARIN Board of Directors adopt: 2008-6: Emergency Transfer Policy for IPv4 Addresses http://www.arin.net/policy/proposals/2008_6.html LACNIC is currently discussing a transfer proposal: LAC-2009-04 Transfer of IPv4 Blocks within the LACNIC Region http://www.lacnic.net/documentos/politicas/LAC-2009-04-propuesta-en.pdf 4. Details of the proposal ---------------------------- APNIC will process IPv4 address transfer requests involving current account holders following the adoption of this proposed policy, subject to the following conditions: 4.1 The minimum transfer size accepted will be a /24. 4.2 APNIC is to maintain a public log of all transfers. 4.3 Address transfers should be permitted between APNIC account holders and NIR members, if and when individual NIRs implement the transfer policy. 4.4 Address transfers are permitted between APNIC account holders and other RIR account holders, following the policies of all the respective RIRs. 4.5 This proposal to take effect as soon as the APNIC Secretariat can implement the mechanisms of the policy. 5. Advantages and disadvantages of the proposal ------------------------------------------------- 5.1 Advantages This proposal: - Ensures that the APNIC registry continues to reflect the current actual status of IPv4 resource holdings by APNIC account holders. - Mitigates the risks to the integrity of the network, and its routing and addressing infrastructure in particular, associated with the unregistered transfers of IPv4 addresses. - Provides a stronger incentive for unused IPv4 address space to return to active use, helping to satisfy residual demand for IPv4 address space across the IPv6 transition. 5.2 Disadvantages (and responses) 5.2.1 Altering the traditional concepts of IP addresses This proposal has the potential to alter a number of traditional preconceptions relating to addresses and their value, including challenging the concept that addresses are not in and of themselves assets and addresses do no in and of themselves have monetary value outside of the narrow constraints of use in networks for routing and end point identification. Changing these common percpetions about addresses and their use opens up the potential for a number of responses, including: - The loss of strong aggregation capability in the address space, with the consequent load being imposed on the routing system. - The significant shift away from a universal need-based address allocation model in the underlying policy framework. - The treatment of addresses as property with the associated legal ramifications in terms of corporate and contract law. - The imposition of taxes on addresses and their movement. - The potential for unfairness and inequities to be realized in terms of access to addresses for use by network service providers. Response: A number of factors mitigate the risks above: - As the transition to IPv6 gathers pace, any residual value of IPv4 addresses would fall in line with the decreasing value proposition of IPv4-based services in an increasing IPv6 network. - If this policy were to be adopted while IPv4 addresses are still available from APNIC, APNIC's established IPv4 address allocation process would continue to provide an alternative source of supply of IPv4 addresses to the industry. 5.2.2 Proposal diverts attention from address reclamation and resuse It has been argued that the proposal diverts attention from policy development that encourages IP address reclamation and reuse. Response: To date the level of return and reclamation of addresses has been minimal. Aside from price-based mechanisms it is unclear that further policy refinement would alter this situation. Even if policy development encouraged address reclamation and reuse, there is the distinct possibility that the amount reclaimed addresses would be smaller than the amount needed for APNIC to continue to allocate addresses on a needs-basis after the unallocated address pool has been exhausted. An open and significant issue is how APNIC could fairly ration limited addresses when faced with a much larger set of demands. In other words, concentrating on relamation and reuse policies rather than transfer policies also contains significant issues that may prove challenging to resolve as a policy matter. 5.2.3 Potential for APNIC to be cast as a regulator If APNIC adopted this policy, APNIC may be cast as a regulator of a secondary market in addresses. Concerns have been expressed that APNIC has no experience, expertise nor the authority to enforce regulatory actions. Such a role may also expose APNIC to additional litigation. Response: This proposal does not advocate such a role for APNIC. The scope of this policy is explicitly limited to the conditions that would allow APNIC to recognise and record a transfer of addresses in its registry. There is a general belief that adoption of this policy would act as an incentive for a market in addresses. However, that does not imply that markets would act outside existing regulatory structures. Nor does it mean that market participants would be immune from existing regulatory measures within their respective regimes. The potential for additional liabilities associated with this policy should be the subject of legal review by an appropriately qualified party. 5.3 Summary of comments on transfer proposals There are a number of views of this that have been noted in the various policy discussions on this topic in the various RIR policy forums. The APNIC policy proposal is broadly similar to a policy proposal under consideration in RIPE, which is referred to here as a "minimal' policy for address transfers. The address transfer proposal currently under consideration in ARIN has a larger set of constraints to be applied to determine if a transfer would be recognised by the registry. A summary of the discussion of the differences in these policy proposals follows. 5.3.1 In favor of a 'minimal' policy - The policy places APNIC in the role of a 'title office' for addresses, and ensures that APNIC, as a registry operator, is neutral in terms of the means used by APNIC members to determine that they wish to proceed with a transfer of addresses. As long as the criteria for a valid transfer are met, by whatever means agreed to by the parties involved, then the registry should allow the transaction to be duly recorded. - APNIC has no practical operational experience in the area of enforcing various constraints on parties as to how and why addresses may be transferred, and does not currently have any recognized authority to do so. Making policy in the absence of a well understood and commonly accepted authority model calls into question the legitimacy and authority of outcome. - Regulation is a well understood and familiar concept in many, if not all, regimes. If there is a general desire to place constraint and regulate the actions of parties who wish to undertake a transfer of addresses, then it may be preferred to do so in the context of a broader framework that involves other bodies and authorities that have a greater level of experience and authority in this area of activity, and leverage from existing regulatory structures and enforcement mechanisms. In this manner the policy proposal does not attempt to create a novel, and potentially superfluous additional regulatory framework. - APNIC has no experience in determining what actions by potential parties to a transfer may need to be constrained in some fashion. Attempting to create policy in anticipation of the need for such constraints is going to be a guessing game that has accompanying flaws, Irrespective of what constraints are initially specified in policy, it will be the case that as the levels of experience in this form of activity increases some iterations over the policy of constraints will be necessary in any case. This approach argues to start from a position that is relatively open and unrestricted, and recommend that APNIC impose additional constraints only when all other forms of constraint are inapplicable and there is a clear need and common desire for such constraints to be enforced by APNIC as distinct from using another party for such a role. 5.3.2 In favour of applying a greater level on constraint in the policy - APNIC has no practical operational experience in address transfers, so we should take incremental steps form the current position rather than a complete relaxation of the entire set of constraints associated with the existing allocation framework. Recipients of a transfer should be qualified by the registry on the basis of demonstrated need in the same fashion as recipient of address allocations today. Address blocks should not be arbitrarily fragmented. Timing constraints should be applied to stop transfers of addresses occurring that are primarily motivated by reasons other than immediate need for use in deployed networks. - Constraints that are generally considered to be onerous and unnecessary can always be removed at a later date, while applying and enforcing additional constraints at a later date will prove to be a far harder task. - There are a number of technical risks associated with address trading. Unconstrained deaggregation will lead to a fracture of the routing system due to unconstrained and large scale expansion of the inter-domain routing table. This is an irreversible state. 5.3.3 Discussions of the emergence of a market The various comments made on this and the related address transfer policy proposals provides grounds for the observation that there is a general perception that the recognition of address transfers leads to a de facto recognition of the emergence of a market or markets for IPv4 addresses. A related topic of discussion about the merits or otherwise of these proposals has been the consideration of the relative merits and risks of market behaviours when applied to this situation. The comments opposed to the emergence of markets include the following: - Markets in addresses are an inevitable consequence of a transfer policy, and unconstrained markets are prone to a number of risks of distortion. These risks include deceptive trading, margin trading, trading in market derivatives and futures, hoarding, and speculation. The utility of an address as a token for addressing a packet is devalued in a market situation. - Operation of a market will lock out all but the largest of players in the network from access to further addresses, as the value of the address will be set by the bidder with the highest price and the ability to exploit the address to its maximal extent. - The operation of a market in IPv4 addresses will lead to the erosion of the effectiveness of self-imposed policies in IPv6, and may lead to the emergence of a market in IPv6 addresses. - Markets are unfair in terms of the implicit bias of a market in favour of those parties who are in a position to set the market price, and inherently discriminating against those parties who do not have the capacity to pay. In an international context this is counter to the general objective of a generally available, neutral and non-discriminatory communications infrastructure. Comments in favour of the emergence of a market in IPv4 addresses include: - Address exhaustion poses an insoluble problem to the address registries in that for as long as there is a continuing demand for addresses the registries have no means to meet that demand. Markets create a means for addresses to be recycled, and create a means for the various levels of demand and supply of addresses in IPv4 to reach a balance through a market-based pricing mechanism. - At every stage there is always an alternative to bidding for IPv4 addresses in the context of a market transaction: namely the use of IPv6 within the network, and the use of an upstream protocol translation service to provide legacy access to other IPv4 networks. Given that substitutes exist, the potential price of IPv4 addresses in a market is capped by the cost of deployment of IPv6 and IPv4 transitional mechanisms. - This is a temporary measure during the dual stack phase of IPv6 transition. The higher the market price for IPv4 addresses the greater the cost pressure placed on the industry to undertake the IPv6 transition, which in turn limits the lifetime of the market and the speculative potential of any such market. Players will have an incentive to act quickly in terms of releasing address resources into such a market given that withholding for too long will result in no return as the market will naturally terminate once IPv6 transition has reached a critical deployment mass. It should be noted that this address transfer policy proposal is mute on the topic of a market for address transfers, and neither advocates nor specifically opposes the emergence of any such market or markets. The policy constrains itself to enumeration of the set of constraints that would apply for APNIC to recognise and register a transfer of addresses between APNIC members. How those parties arrived at the decision to undertake the transfer, and the related issues concerning property, financial and legal frameworks and the emergence of markets, the need to regulate such markets and identification of the market regulator are specifically not encompassed by this policy proposal, nor does this proposal advocate that such a role be undertaken by APNIC. 6. Effect on APNIC members ---------------------------- APNIC members will have the ability to register the transfer of IPv4 address resources between APNIC members. 7. Effect on NIRs ------------------- As stated in section 4.3 of this proposal, NIRs will have the ability to register transfers of addresses with APNIC account holders if and when individual NIRs implement the transfer policy.