Summary

  • APNIC designed prop-062 in 2008 and implemented it on 16 February 2009, before the ration was needed. When the regional pool reached its final /8 on 15 April 2011, every new or existing account holder could obtain no more than a /22, or 1,024 addresses, from 103/8 subject to eligibility. The quota was attached to an account, not automatically to a corporate group, operating network or beneficial controller.
  • The rule achieved its central conservation result. APNIC delegated from 103/8 for more than 12 years. A 2019 reduction from /22 to /23 cut the maximum to 512 addresses, and APNIC later estimated that the change extended the block's life by about two years. On 9 October 2023, APNIC reported only 121 /24 equivalents left and expected the final 103/8 delegations that week.
  • Applicant behavior changed immediately because an account carried an option on below-market IPv4. APNIC reported in 2013 that new members had been created to receive final /8 allocations and then quickly transfer them to an existing member. At that point, 32 of 2,110 final /8 delegations had transferred, 15 members held more than a /22 from 103/8 and one held five /22s. These observations were limited in scale but directly contradicted the assumption that one account always represented one independent entrant.
  • Policy adjustment lagged the observed behavior. A 2013 proposal to restrict excessive transfers from 103/8 was abandoned. A later rule, implemented in November 2017, imposed a five-year transfer restriction. By then APNIC reported that 175 members had accumulated more than /22 through transfers or mergers and acquisitions, while 31% of direct and National Internet Registry members had not received a final /8 delegation.
  • Rationing displaced rather than eliminated demand. Applicants used recovered-pool allocations, market transfers, provider-assigned space, leasing, carrier-grade NAT, cloud addresses and IPv6. Prop-105 created a second potential /22 from recovered space in 2014; in 2019 the pools were consolidated and the 103/8 maximum was reduced. Each adjustment changed the relative value of accounts, transfers and technical substitutes.
  • Fairness cannot be inferred from the number of account holders served. A proper evaluation needs account-to-controller mapping, new-entrant status, transfer and merger history, routed use, downstream customers, processing time, fees, rejected requests and substitution costs. The experiment preserved a broad minimum opportunity, but it also rewarded actors able to create, acquire or coordinate multiple qualifying entities.

The scarce unit was not the address; it was the qualifying account

APNIC's final /8 policy is often summarized in one sentence: each account holder could receive a /22 from 103/8. That sentence contains the experiment's most important design choice. The quota did not attach to each person in the Asia Pacific, each autonomous system, each corporate group, each network or each economy. It attached to the holder of an account that met APNIC's criteria.

A /22 contains 1,024 IPv4 addresses. It is tiny beside the needs of a national mobile operator, a large hosting platform or a mature broadband network. It can nevertheless be valuable to a new network that needs public infrastructure, translation capacity, authoritative services or a small pool of customer addresses. The policy tried to reserve that foothold for both current and future entrants instead of letting the remaining block disappear into the largest contemporary requests.

The conservation logic was strong. If a /8 contains 16,777,216 addresses, dividing it into bounded portions can support many more recipients than continuing ordinary needs-based allocations to large networks. Reserving space for unforeseen uses reduced the distributable total, and allocation geometry created overhead, but the order of magnitude remains clear. A fixed maximum converts a rapidly vanishing stock into a long-lived queue of small opportunities.

The account boundary, however, is an administrative identity. A commercial group can contain several legal entities. A network can use accounts through a National Internet Registry as well as direct APNIC relationships. A member can be acquired. Separate infrastructures can share owners, and apparently independent applicants can have common control. Conversely, one account can stand behind many downstream operators and customers.

Once a ration has value, the boundary that earns it attracts organization. That does not mean every new entity is artificial. APNIC's region experienced genuine growth in new networks, especially in developing economies. It means the rule cannot assume that counting accounts is the same as counting independent entrants. The policy turned membership and entity definition into part of IPv4 economics.

The ration was designed before the emergency arrived

Prop-062 was submitted in July 2008, reached consensus at APNIC 26 in August and was implemented on 16 February 2009. The design therefore preceded the trigger by more than two years. It was not an improvised response to an empty shelf.

The proposal tied the final /8 distribution size to APNIC's minimum allocation size and applied the opportunity to existing and future account holders. The design goal was continuity for newcomers in a world where an entirely IPv6-only network could not yet communicate smoothly with the IPv4 Internet. A small IPv4 block was treated as a bridge rather than a growth inventory.

On 3 February 2011, IANA allocated 103/8 to APNIC as one of the final five /8 blocks distributed to the Regional Internet Registries. APNIC announced that its regional pool reached the final /8 stage on 15 April. Delegations from 103/8 began on 18 April. From then, ordinary large allocation logic gave way to the maximum /22 entitlement.

APNIC Executive Council minutes around the transition recorded 2,635 members in March 2011 and 180 requests in the queue when the final stage arrived. Those figures show that the policy did not enter an empty administrative environment. Applicants were already waiting, and the registry was serving a rapidly growing region.

The advance design was a governance strength. It reduced the risk that the last large applicants would consume the block before a public rule existed. It gave operators notice and allowed APNIC systems to prepare. Yet advance design necessarily rested on assumptions about future membership, IPv6 maturity, address demand and the meaning of an account. The following decade became a test of those assumptions.

A /22 was a bridge for one network and a rounding error for another

Equal rations are not equal economic outcomes. A /22 can support several public services, a modest hosting footprint, NAT gateways or the early customer base of a small provider. It cannot supply one public address per subscriber to a growing mobile or fixed network. The same allocation can be transformational for a new entrant and operationally negligible for an incumbent.

That divergence forced substitution. A small operator could stretch 1,024 addresses through careful assignment and address sharing. A large operator needed transfers, leasing, upstream provider space, carrier-grade NAT or a faster IPv6 transition. A cloud or hosting company might combine all of them. The quota did not cap demand. It capped one channel of supply.

This is why the ration acquired a shadow price even though APNIC did not auction it. The holder paid membership and service fees rather than the market value of the addresses. If a transferable /22 could command more than the cost of forming and maintaining an account, the right to apply carried economic value. If transfers were restricted, the ration still saved the holder from buying the same quantity elsewhere and could support a revenue-generating network.

The gap between administered cost and market value encouraged legitimate entry and strategic behavior at the same time. A genuine entrant had a lower initial barrier. An established group had an incentive to ask whether a separate qualifying entity could obtain another ration. A holder that no longer needed its block had an incentive to sell, merge or reorganize rather than return it without consideration.

No rule can remove these incentives by declaring addresses non-commercial. It can alter the cost, timing and legality of each response. The 103/8 experiment is valuable because its later policy history records those responses in unusually direct terms.

Applicants adapted the size and sequence of requests

Rationing changed the request itself. Under ordinary allocation, an applicant tried to demonstrate a block sized to forecast need. Under the final /8 rule, a qualifying account knew the ceiling in advance. The rational request was often the maximum it could justify up to that ceiling, even if short-term need was lower, because the account could not return later for unlimited growth from the same pool.

Policy changes allowed delegations to be made in more than one request until the account reached its total maximum. That improved fit for organizations needing less than a /22 initially. It also created timing choices. An applicant could take a /24, preserve room for later requests and watch policy or demand change. The registry had to track cumulative entitlement rather than a single allocation.

The ceiling simplified some decisions and complicated others. Staff no longer had to choose between a /16 and a /18 from 103/8, but still had to determine eligibility and need. Applicants had to decide whether small fragments were worth the routing and management cost. A /22 ceiling could arrive as one block or several smaller prefixes, which are not economically identical.

The predictable maximum also influenced application timing. An operator that expected the pool to last had less reason for a panic request than under an unrestricted countdown. But a change in transfer rules, fees or the expected exhaustion date could accelerate claims. Every account that delayed faced the possibility that later policy would reduce the maximum, as occurred in 2019.

This is the first behavioral lesson: rationing replaces a quantity contest with an eligibility and timing contest. It is calmer than a run on the pool, but it does not eliminate strategy.

APNIC saw account multiplication and consolidation by 2013

The strongest evidence of induced entity behavior comes from APNIC's own transfer reporting. At APNIC 35 in February 2013, staff said they had observed new members being created to receive final /8 allocations and then quickly transferring them to an existing member, possibly a parent company. A presentation reported 32 transferred delegations out of 2,110 from 103/8, or 1.5%.

The same report said 15 APNIC members held more than a /22 from 103/8 and that the highest observed account held five /22s. It also described a group of companies transferring blocks among themselves in circles, while carefully calling the suspected fee motive a suspicion rather than a proven conclusion.

These numbers require proportion. A 1.5% transfer rate did not show that most final /8 recipients were gaming the rule. Staff noted that it was close to the 1.8% transfer share among earlier delegations. The evidence did show that the one-account ceiling was permeable through later transfers and corporate relationships. A rule designed to limit each account's initial receipt did not necessarily limit the amount eventually consolidated under one account.

The observed sequence matters more than the share. Create an account, receive the ration, transfer it to an existing account and close or reduce the original relationship. That sequence treats the account as an acquisition vehicle. It converts administrative identity into a temporary claim on scarce stock.

This behavior was not necessarily prohibited at the time. APNIC's 2016 explanation of the issue said new member organizations could apply and then transfer without breaking existing policy. The problem was a gap between rule text and stated purpose. The rule limited initial delegation per account, while the transfer policy allowed later consolidation.

The experiment therefore exposed a basic design principle: if a quota is intended to apply to an economic group, the policy must define and evidence common control. If it applies only to an account, consolidation is not an anomaly; it is a foreseeable response.

The first anti-consolidation proposal failed

Prop-106, submitted in January 2013, directly addressed the behavior. Its synopsis said some Local Internet Registries appeared to collect final /8 blocks through multiple accounts and transfer them to a single account. The proposal sought restrictions on excessive transfers from 103/8. It failed to reach consensus at APNIC 35 and was abandoned on 28 February 2013.

Failure did not prove the observation false. It showed that the institution had not agreed on the remedy. Entities could reasonably worry about retroactive restrictions, legitimate business changes, address efficiency, transaction rights and the difficulty of identifying abusive common control. A transfer can consolidate an artificial structure, but it can also move addresses from a failed entrant to an operator that will use them.

The policy choice left the initial-account rule and general transfer rule misaligned for several more years. During that period, an organization could lawfully receive the ration and later move it, while the pool's stated purpose remained long-term entry support.

This lag is central to the experiment. A policy can observe strategic response quickly yet take years to produce an accepted correction. The delay is not just procedural. It changes incentives for every applicant acting under the existing rule. Early actors receive options that later actors may not.

The abandoned proposal also reveals the limits of moral language. Calling consolidation contrary to the spirit of the final /8 did not define a workable legal or operational test. Was common ownership enough? What if infrastructures remained separate? What if a group acquired an entrant after years of genuine operation? What if transfer prevented addresses from becoming idle? The rule needed a mechanism, not an intention.

Corporate form became part of address acquisition

Any per-entity ration creates pressure on entity definition. Corporate groups routinely separate businesses for licensing, tax, investment, risk and local regulation. The registry must distinguish genuine operational separation from a shell created mainly to claim another ration, but the same documents can support both stories.

APNIC's current merger and acquisition policy illustrates the difficulty. It allows a membership to continue when an entity becomes a subsidiary and the infrastructures remain fully independent. If infrastructures merge, APNIC will not continue separate allocations to both. The policy requires disclosure of address holdings across the relevant entities and permits review of the practical effect of ownership change.

That practical-effects test is more economically meaningful than legal names alone. Separate companies under common ownership may operate distinct networks with separate customers, staff and routing. Conversely, two formally independent applicants may be coordinated by one controller. Corporate registry extracts cannot answer the network question by themselves.

The test is also expensive. Applicants may need ownership charts, acquisition documents, network diagrams and utilization records. APNIC must interpret structures across many jurisdictions and languages. National Internet Registry relationships add another layer. Small legitimate groups can face the same evidence burden as deliberate account multiplication.

The quota therefore changes company architecture at the margin. It raises the value of keeping infrastructures and memberships separate after acquisition. It can make a merger of networks costlier if consolidation threatens separate address entitlements. It can encourage a group to preserve subsidiaries that would otherwise integrate. Those are real organizational effects produced by an address rule.

They should not be overstated. Companies form subsidiaries for many reasons, and public data rarely prove that IPv4 was decisive. The evidence supports a narrower claim: APNIC observed account creation and transfer behavior consistent with quota response, and its later policies had to examine ownership and infrastructure separation because legal form alone was not enough.

The recovered pool created a second ration and a new substitution

Prop-105, implemented in May 2014, allowed each APNIC account holder to apply for an additional /22 from a separate recovered pool. This altered the experiment. An account that had exhausted its 103/8 entitlement could potentially receive another ration from returned or post-exhaustion IANA space.

APNIC's March 2015 notice about a new IANA /13 made the distinction explicit: qualifying account holders could apply for up to a /22 from the recovered pool and up to a /22 from 103/8, provided they had not already claimed from the relevant pool. The practical ceiling through APNIC could therefore reach a /21 across two channels for an eligible account, although availability and rules differed.

This was not a hidden loophole. It was an adopted allocation choice about recovered stock. It nevertheless changed behavior. The value of membership increased because one account could hold claims on two pools. Applicants had to track separate eligibility. Demand that would otherwise move to the transfer market returned temporarily to administered supply.

APNIC's ten-year review says recovered-pool distribution caused IPv4 delegations to rise from 2014 to 2016, after which that pool was quickly exhausted and a waiting list formed. The substitution was temporary because the recovered stock was small relative to regional demand.

The two-pool design also complicated fairness claims. An account count served from 103/8 did not show whether the same account received recovered space. A new entrant with one ration and an incumbent with entitlements from both pools were both counted as members but held different address options.

In July 2019, APNIC abolished the non-103/8 waiting list and treated recovered non-103 space under the same remaining-pool policy. Consolidation simplified administration but changed the economic meaning of recovery. Returned addresses no longer recreated the earlier allocation regime; they fed the rationed channel.

Transfers became the pressure valve and the escape route

A fixed ration only works as an entry guarantee if networks needing more can obtain supply elsewhere. In the APNIC region, transfers became the principal market pressure valve. APNIC's policy permitted unused addresses to move between account holders and later across compatible registry regions, subject to recipient conditions and accurate registration.

The transfer market served several functions at once. It supplied large operators whose requirements dwarfed /22. It let holders realize value from unused legacy or allocated space. It allowed a growing entrant to move beyond the ration. It also enabled consolidation of 103/8 blocks in ways that weakened the original per-account distribution pattern.

By APNIC 42 in 2016, staff reported that 175 members had accumulated more than /22 through transfers or mergers and acquisitions. In the same discussion, 31% of direct APNIC and National Internet Registry members had not received a final /8 delegation. Less than half of 103/8 remained.

Those statistics should not be combined into an accusation. A member with more than /22 may have bought legitimate supply, acquired another company or inherited separate networks. A member without 103/8 may not have needed it. The comparison demonstrates that the initial ration did not determine final holdings. Market and corporate events redistributed the stock.

That redistribution is not inherently a policy failure. A ration should not freeze addresses in an unsuccessful firm while another network has customers. The governance question is whether transfer rules preserve the entry purpose long enough to matter without trapping underused resources. APNIC first favored movement, then moved toward a holding period.

The five-year lock arrived after the market had learned the rule

Prop-116 was first discussed in 2016 and reached consensus at APNIC 44 in September 2017 after amendment. The community extended the proposed holding period from two years to five. APNIC implemented the rule on 20 November 2017. Current policy prevents 103/8 addresses from being transferred for at least five years after their original delegation, including through mergers and acquisitions.

The rule changed the option embedded in a new account. A recipient could still use its ration, but could not quickly convert it into a recognized transfer. That raised the cost of account multiplication for immediate sale and better aligned the initial delegation with operating use over time.

It also imposed costs on legitimate change. A startup can fail within five years. A corporate group can reorganize. An operator can discover that its original need no longer exists. Current policy says that if the original reason is no longer valid during the restricted period, the addresses should return to APNIC rather than move for consideration. That protects the rationed pool but assigns the residual value away from the holder.

Retroactivity became contentious. Prop-123 in 2018 sought to exempt delegations made before 14 September 2017 from the five-year restriction. The proposal did not reach consensus and was withdrawn. A mailing-list discussion said 10,868 103/8 delegations had occurred during the five-year lookback period before that debate, illustrating how many prior decisions could be affected.

The holding period therefore solved one form of arbitrage by creating another timing boundary. A block just outside five years became transferable; a similar block one day inside did not. Markets price such dates. The restriction changed liquidity without changing technical usability.

This is the second major lesson: delayed anti-gaming rules create cohorts. Early applicants operate under one option set, later applicants under another, and transitional disputes become part of scarcity economics.

Cutting the ration to /23 conserved stock but doubled the shortfall

On 28 February 2019, prop-127 reduced the maximum 103/8 delegation from /22 to /23. The account entitlement fell from 1,024 addresses to 512. APNIC's 2019 Annual Report says the change effectively halved annual consumption compared with the previous year. APNIC later estimated that it extended the life of 103/8 by about two years.

The conservation effect is straightforward. Given the same number of qualifying accounts and similar take-up, halving the maximum roughly doubles the number of rations the remaining stock can support. The policy preserved access further into the future.

The economic effect is equally straightforward. A /23 leaves less room for public services, translation, customers and growth. Applicants need transfers, leases, provider space or sharing sooner. The ration's value per address may rise because it remains below market acquisition cost, but its value as a self-contained operating block falls.

The reduction also changed cohort fairness. Earlier accounts could hold a /22 from 103/8, while later accounts could obtain only /23. Some earlier holders had also accessed the recovered pool before consolidation. The policy preserved future entry by accepting unequal entitlements across time.

That trade-off may be defensible, but it should be measured. How long did the added two years support genuinely new operating networks? How many later /23 recipients bought additional space within 12 or 24 months? How many used carrier-grade NAT or upstream addresses? Without these outcomes, longevity alone can make a ration look successful even if the amount ceased to be operationally sufficient.

The correct denominator is not simply years of pool life. It is entrant-years of useful connectivity produced per address withheld from larger demand.

Rationing pushed costs into network architecture

When an operator cannot obtain enough public IPv4 through APNIC, the shortage appears elsewhere. Carrier-grade NAT allows many customers to share addresses but adds equipment, logging, port management, abuse attribution and support complexity. Provider-assigned space lowers initial acquisition cost but increases renumbering and supplier dependence. Cloud public addresses and managed NAT convert scarcity into recurring platform charges.

Leasing can provide flexible capacity without an immediate purchase, but the operator depends on contract renewal, clean reputation and the lessor's continuing authority. Market transfers supply recognized control at a price and with diligence. Corporate acquisition can deliver address holdings but brings business and legal complexity. IPv6 reduces dependence where counterparties and applications support it, yet dual-stack operation often remains necessary.

These substitutes differ in who bears risk. The quota made APNIC's pool last longer by limiting one visible allocation. It did not make the region's underlying demand disappear. Costs moved to operators, customers, upstreams, cloud platforms, brokers and compliance teams.

This displacement complicates claims that rationing was free or universally fair. A small entrant received a valuable foothold, but one that might require immediate spending on NAT or additional addresses. A large incumbent with existing stock could wait, optimize or buy at scale. A new low-margin ISP faced address-market prices before earning the customers needed to finance them.

The policy should therefore be judged as an insurance design, not a full supply policy. It insured that many qualifying accounts could receive something. It did not insure that the something was enough to compete on equal terms.

National Internet Registries blurred the denominator further

APNIC's region includes National Internet Registries that mediate resource relationships in several economies. The final /8 policy applied to APNIC and NIR members, but an NIR member, a direct APNIC account and a corporate group are different units.

An account-based report can count one direct member serving a large multinational network and one small local operator as equal recipients. An NIR can aggregate many downstream members behind its relationship with APNIC. Policies and fees at the national layer can affect application timing and evidence. A group can have entities in several economies and interact with different registration channels.

This does not mean NIRs caused circumvention. It means a regional fairness analysis must trace the delegation chain. APNIC should be able to distinguish direct accounts, NIR accounts, downstream recipients, common controllers and actual routed networks without publishing confidential corporate detail.

The NIR layer also affects substitution. A transfer between an APNIC account and an NIR member can require coordination across institutions. Local fees, language and company evidence influence transaction cost. An address ration that is uniform in regional policy can feel different in national execution.

Counting allocations by economy adds another risk. A country with many new memberships may reflect genuine network formation, strategic account creation or both. APNIC's own reviews identify strong membership growth in Bangladesh, Thailand, Myanmar and Mongolia. Those trends are important evidence of entry, but membership alone cannot establish that the final /8 caused sustainable competition.

The denominator must move from accounts to operating outcomes: independent networks connected, customers served, duration of use, transfers, closures, common control and the cost of supplementary supply.

Routing data can test use, but not motive

Public delegation and routing data provide a way to examine whether 103/8 blocks became visible in BGP, how quickly they were announced and whether origins changed after transfer. APNIC research in 2015 examined the block and found thousands of entities had received final /8 allocations. It also identified registration anomalies and likely post-allocation mergers.

Routing evidence can show that a block is announced, but not whether the applicant was a genuine entrant, a subsidiary or a consolidation vehicle. A parent and child can use the same upstream. A leased block can route from a customer's autonomous system while remaining registered elsewhere. An unannounced block can support private interconnection or be reserved for later deployment.

Changes in origin are similarly ambiguous. They may indicate transfer, transit migration, DDoS protection, multihoming or a route leak. Registry logs and corporate records are needed to interpret them. APNIC's 2013 transfer figures are stronger evidence of recognized consolidation than BGP inference alone.

A rigorous evaluation would combine delegation dates, account status, transfer logs, merger records, routing origins and controlled common-ownership identifiers. It would publish aggregates, not expose customer networks. That would allow analysts to distinguish allocation longevity from useful deployment and later concentration.

The absence of such a linked public series is a material limitation. It means strong conclusions about the prevalence of gaming would exceed the evidence. The known cases prove the mechanism existed; they do not prove it dominated the pool.

The pool lasted, but longevity is only one outcome

By October 2023, APNIC was nearing the end of ordinary delegations from 103/8. The registry reported that only 121 /24 equivalents remained available in that block at the time of writing. It also reported almost 20,000 /24 equivalents across available and reserved IPv4 pools outside and inside 103/8, much of it associated with recovered space and specific reservations. The end of 103/8 did not mean every APNIC-administered IPv4 address had vanished.

The twelve-year life exceeded the roughly ten-year expectation recalled in APNIC's retrospective. On the narrow conservation measure, prop-062 and later reductions worked. Many cohorts of entrants retained access to a small delegation long after ordinary allocation ended.

Longevity cannot answer who captured the benefit. Each year of life is valuable only if the recipients were networks the policy intended to support and if the allocation improved their ability to connect. Transfers, mergers, closures and common control can move the benefit after initial delegation. A /23 received in 2022 is not equivalent to a /22 received in 2012.

Nor can longevity measure the cost imposed on larger demand. Addresses denied to one operator may have supported more customers if allocated differently. That does not make large allocation socially superior; it shows that rationing chose breadth of recipient access over immediate scale. The choice deserves explicit reporting rather than a generic statement of fairness.

The experiment succeeded at preserving the option. Whether it maximized entry, competition or user welfare is a separate empirical question.

Recovery changed scarcity without restoring the old world

APNIC's historical-resource transition, completed in 2023, illustrates how recovered stock can reopen supply after a final /8 runs down. APNIC reported 658,944 IPv4 addresses recycled, 3,705,344 retained by custodians and 2,113,536 reserved during the project, with some cases still unresolved.

Recovered addresses are not identical to untouched free-pool inventory. Their authority history may be complicated. They may retain route, reputation and geolocation memory. Some are reserved while claims are investigated. Returning them to use requires notice, evidence and operational cleanup.

The current single-pool approach treats recovered and later IANA-supplied space under the remaining ration rules. That preserves access but keeps the account boundary central. Every recovered block renews the question: should it support new entrants broadly, relieve a waiting list, satisfy larger demonstrated demand or enter market circulation through holders?

Recovery also creates expectations that can affect applicant timing. A network may delay a transfer in hope of receiving administered space. A five-year transfer restriction can return some blocks to APNIC when original need ends. Enforcement choices therefore influence supply and price.

The pool is no longer a one-way depletion clock. It is a stock with small inflows from returns, recovery and global distribution, set against continuing demand. Rationing remains a policy choice about those inflows, not an automatic consequence of protocol scarcity.

Fairness requires five denominators, not one success statistic

An account-holder count measures administrative reach. It does not by itself measure fair distribution. At least five denominators are needed.

First is the account denominator: how many eligible direct and NIR-linked accounts received a delegation, and how many remained unserved? Second is the controller denominator: how many independent beneficial or operational groups controlled those accounts at application and later? Third is the network denominator: how many distinct operating networks put the space into use? Fourth is the customer denominator: how many users or services depended on those networks? Fifth is the time denominator: how long did the original entrant retain and use the block before transfer, merger, closure or return?

Each answers a different question. A thousand accounts under common control would perform well on the first and poorly on the second. One large operator serving millions may perform poorly on recipient breadth and strongly on customer reach. A block transferred after a genuine startup failure can still produce long-run use even though original-entrant retention is low.

Policy should not collapse these measures into a moral ranking. It should publish them so the distributional choice is visible. Breadth, independence, use, reach and durability can conflict.

The same report should include costs: membership and sign-up fees, median APNIC-controlled processing time, evidence burden, transfer prices where voluntarily available, NAT expense proxies and the share of recipients acquiring additional IPv4 within two years. Without cost, a small allocation can be counted as access even when it is too expensive to make operational.

A better quota would follow control, use and exit

If APNIC retains rationing for recovered supply, the rule can learn from 103/8. The eligibility unit should be explicit. If the cap is per account, policy should admit that related entities can each qualify when they operate genuinely separate networks. If the cap is per controlling group, APNIC needs a proportionate common-control test and an appeal route.

The test should rely on practical indicators: shared ownership, management, network operations, customers, facilities, routing policy and finances. No single indicator should decide. Small applicants should not face an investigation more expensive than the ration. Random or risk-based review can deter artificial splitting without treating every subsidiary as suspect.

Transfer restrictions should follow a stated purpose and include humane exits. A cooling-off period can deter immediate resale, but a failed entrant should have a transparent route to merger, return or supervised transfer. The rule should distinguish distress, genuine business combination and planned arbitrage. Decisions need reasons and timely review because a delay can destroy remaining enterprise value.

APNIC should publish cohort outcomes. For each delegation year and size, report active account status, transfer or merger incidence, return, routed visibility and common-control concentration in aggregate. Compare /22 and /23 cohorts. Report how many recipients obtained additional supply through transfer or recovered pools.

Such reporting would turn the final /8 from an institutional story into a testable policy experiment. It would also let future rules change before a known response becomes entrenched for years.

The registry should record the market it creates at the margin

APNIC did not set a cash price for 103/8, but the quota changed market prices and choices. Every below-market ration reduced one recipient's demand for transfers. Every reduction from /22 to /23 increased likely supplementary demand. Every transfer lock reduced short-term supply. Every recovered-pool distribution shifted buyers away from private acquisition.

This is not an argument that APNIC should control prices. It is an argument that policy has market effects even when written as conservation. The registry should measure those effects as part of accountability.

Useful statistics include transfer volume and block size by 103/8 cohort; time between delegation and transfer; transfer incidence before and after the five-year rule; number of accounts later linked by common control; membership closures after transfer; and concentration of final /8 holdings after mergers. Price data can remain voluntary and aggregated.

The report should also include substitution: applications for IPv6 alongside IPv4, growth in transfer pre-approval, requests for provider-independent space, and surveys of NAT or provider-assigned dependence. No single metric proves causation, but a consistent series can show how behavior moves when the quota changes.

Rationing is often defended by pointing to the remaining pool. Market accountability asks what changed outside the pool. Both views are necessary.

The experiment did not fail; it revealed the price of its boundary

It would be wrong to call prop-062 a failure simply because applicants adapted. All scarcity rules induce adaptation. The policy prevented a final rush of large allocations and kept a minimum IPv4 opportunity available to new and existing accounts for more than a decade. That was a substantial operational result.

It would be equally wrong to call the experiment complete because the block lasted. The account-based ceiling created an asset-like option. APNIC observed new memberships followed by transfers, multiple-account accumulation and concentration beyond /22. Policy took years to impose a holding period. A second recovered-pool ration, a later /23 reduction and pool consolidation repeatedly changed the value of the option.

The policy's strength and weakness came from the same design. A simple account rule was administrable across a vast, diverse region. The simplicity made the account boundary worth organizing around. Tightening the boundary required corporate and operational judgments that were less simple, more intrusive and more costly.

The rational conclusion is not to abolish every quota or to freeze every transfer. It is to acknowledge the trade. Broad minimum access, efficient reuse, corporate freedom and anti-arbitrage control cannot all be maximized simultaneously. Policy must state which objective takes priority and publish evidence of the sacrifice.

103/8 turned institutional identity into scarcity economics

APNIC's final /8 was more than a pool of addresses. It was a long-running experiment in how an administrative rule distributes a valuable option. Prop-062 chose breadth over immediate scale. Prop-105 added a second claim on recovered space. Prop-116 restricted exit for five years. Prop-127 halved the maximum. Pool consolidation changed future recovery. Each decision altered applicant strategy.

The record shows genuine preservation and genuine adaptation. The ration lasted. Membership grew. New networks obtained small blocks. Some applicants also created accounts and transferred allocations. Holdings consolidated through transfers and acquisitions. Operators substituted NAT, leases, provider space, market purchases and IPv6. The rule shaped company boundaries as well as prefix sizes.

The decisive question is therefore not whether APNIC distributed 103/8 fairly in the abstract. It is whether the institution measured the entities and outcomes relevant to its stated aim. An account is easy to count. An independent entrant, a common controller and a sustainable network are harder. That difficulty does not excuse replacing one with another.

Future scarcity policy should preserve the narrow virtue of 103/8: a credible minimum path for a new network. It should also publish control-aware cohorts, recognize legitimate exit, resist rapid account arbitrage and avoid pretending that rationing removes demand. The market and the network will continue to adapt.

The final /8 experiment demonstrates a wider rule of governance. Whenever an institution places a valuable entitlement behind an administrative identity, applicants will optimize both the entitlement and the identity. The quality of the policy depends on whether the institution can see that response, distinguish genuine entry from formal multiplication and adjust before the exception becomes the operating model.

Sources