Summary

  • RIPE proposal 2010-02, accepted in January 2011 and activated when RIPE NCC entered 185/8 in September 2012, limited each LIR to one /22 even when justified need was larger. It also required an IPv6 allocation and reserved a /16 for unforeseen circumstances.
  • The rule preserved a standard first allocation, not a scalable supply. A /22 contains 1,024 addresses; it can support transition, infrastructure or a modest deployment, but its adequacy depends on the operator's service model, sharing design, customer needs and compatibility burden.
  • The policy itself anticipated that organisations could open multiple LIRs and stated that there was then no basis to deny that right. Later RIPE NCC analysis found 5,191 final-/22 allocations by January 2015, with one LIR holding 21 after opening, merging and closing several accounts.
  • Membership growth cannot be read as equivalent to new-firm entry. RIPE NCC's 2019 financial report said LIR accounts rose from 20,624 to 25,125 that year, that impending exhaustion strongly drove the increase, and that about 30% of new accounts from early 2016 were additional accounts.
  • The transfer market became the expansion channel beside the ration. RIPE NCC recorded its first in-region transfer in October 2012 and reported 3,034 IPv4 blocks transferred within the region in 2019. The /22 bought time; it did not remove the need to buy, lease, share or acquire more IPv4.
  • The fair verdict is mixed: the rule prevented immediate capture of the final pool and gave many later LIRs a useful foothold, but it did not preserve equal competitive entry. A better regime would report recipient outcomes, separate legal organisations from accounts, keep transfer recognition narrow and make registry continuity portable.

The policy promised a foothold, not a business model

RIPE NCC's final-/8 policy was unusually candid about its limit. The accepted text of proposal 2010-02 said that an LIR could receive exactly one /22 from the final block even if its demonstrated need justified more. The rationale described the allocation as assistance during the transition to IPv6, not a solution to the growth requirements of larger networks.

That distinction should govern the verdict. A /22 is 1,024 IPv4 address values. It is real, globally routable space and can be operationally valuable. It can support shared-address gateways, public infrastructure, management systems, transition services, a hosting platform with address conservation, or a modest customer deployment. It is not a universal minimum for building a competitive network. Its usefulness varies with architecture.

A mobile operator using large-scale sharing can serve many subscribers behind a comparatively small public pool, though it accepts logging, port-management and troubleshooting costs. A hosting provider offering dedicated addresses may consume the same /22 quickly. An enterprise service provider may need public addresses for customers whose security, remote-access or vendor systems still assume IPv4. A new access network can conserve aggressively and still face customers or upstream services that are not ready for IPv6-only operation.

The policy therefore preserved one kind of entry: access to a recognised first block. It did not preserve every form of commercial entry. It could not guarantee that an entrant would reach efficient scale, satisfy a lender, compete with an incumbent's installed inventory or avoid the transfer market. It was a bridge allocation with a fixed width, while the businesses crossing it carried very different loads.

The official rationale deserves credit for not claiming otherwise. The problem came later when longevity of the pool or growth in LIR accounts could be read as proof that new entry had been saved. Those measures show distribution, not competitive outcome. A ration can reach thousands of accounts while leaving each recipient dependent on a market controlled by historical supply.

The central question is not whether /22 was too small in the abstract. It is whether the policy's objective was defined and measured honestly. If the objective was to ensure that many future LIRs could receive a small routable block, the rule plainly had a mechanical advantage over continued large allocations. If the objective was to preserve entry into IPv4-dependent network markets on comparable terms, the evidence is far weaker.

What changed on 14 September 2012

The regional threshold arrived nineteen months after IANA distributed the final five /8s. On 14 September 2012, RIPE NCC announced that it had begun allocating from 185/8, the final block it had received under the global rule. The announcement described approximately 8,000 members in more than 75 countries and stated that each LIR could receive one /22 after demonstrating need and holding an IPv6 allocation.

The rule changed the relationship between need and quantity. Before the threshold, an accepted requirement could support a larger allocation, subject to prevailing policy and inventory. After the threshold, need remained a condition for entry but stopped determining size above 1,024 addresses. The allocator no longer asked only how much the network could justify. It also imposed a fixed intertemporal reservation for future LIRs.

The arithmetic explains the attraction. A /8 contains 16,777,216 address values, theoretically equal to 16,384 /22 blocks. The accepted policy reserved a /16 for unforeseen circumstances, and practical inventory could include fragmentation, returns and other adjustments, so 16,384 was not a promise of awardable allocations. It was a scale indicator. A fixed /22 prevented a single well-supported large claim from consuming a material share of the remainder.

The institutional unit was the LIR. That was administratively convenient because RIPE NCC already had contractual accounts, contacts, billing and resource records at that level. It also created the rule's central weakness. An LIR account is not necessarily an independent company, a new network, a new competitor or a distinct pool of customers. One legal organisation can operate more than one LIR. An LIR can close or merge. A corporate group can place accounts in separate legal entities. Counting allocations by LIR therefore does not count independent entry.

The policy authors saw the problem. The accepted proposal states that organisations might set up multiple LIR registrations to obtain more space. Its impact analysis went further: current rules allowed one organisation to operate multiple LIRs, and there was no basis at the time for denying that right. The potential circumvention was not an unforeseeable defect discovered years later. It was acknowledged in the design.

The policy nevertheless made a defensible choice under pressure. Defining entitlement by ultimate beneficial owner would have required intrusive corporate mapping across many jurisdictions. Defining it by customer count or revenue would have turned the registry into a business assessor. A standard account-based ration was legible and executable. The cost of that simplicity was that membership structure became a way to acquire additional scarcity.

Four meanings of entry should not be collapsed

An evaluation needs a more precise entry vocabulary.

Registry entry means an organisation establishes or uses a recognised LIR relationship and obtains a first allocation. The /22 rule was designed primarily for this threshold. It created a predictable quantity and kept that quantity available over a longer period than unrestricted claims would have.

Operational entry means the organisation can launch a functioning service. A /22 may be enough, depending on the design. IPv6, address sharing, upstream assignments and carefully limited IPv4 use can stretch it. The block also gives the network direct control over a portable routable range rather than complete dependence on a provider's addresses.

Growth entry means the network can add customers, sites or services without its marginal address cost becoming prohibitive. Here the /22 was explicitly limited public evidence for many models. Once it was consumed, the operator needed transfers, leases, corporate acquisition, provider space, more sharing or a faster shift toward IPv6.

Competitive entry means the entrant can face incumbents without a structural input disadvantage large enough to determine the contest. The rule could not supply this. Incumbents held address portfolios accumulated under earlier allocation conditions. Some had excess stock, some had efficient networks built around it, and some could sell or lease it. The new entrant received a ration and a market problem.

These meanings overlap, but they are not interchangeable. A policy can succeed at registry entry while failing at growth entry. It can improve operational entry for one business model while doing little for another. It can preserve an IPv4 foothold while leaving competitive conditions shaped by historical allocations and capital access.

This framework also prevents an unfair criticism. RIPE NCC could not produce equal competitive entry from a finite final block. Giving every new network enough IPv4 to match large incumbents was mathematically impossible. The legitimate demand is not that the registry abolish scarcity. It is that the institution state which entry it can protect, measure the result and avoid claiming a larger social success than the ration can deliver.

The policy's title concerned allocations from the last /8. Its later reputation concerned fairness to new entrants. The gap between those two is where accountability belongs.

The first outcome test exposed the account loophole

RIPE NCC's own research offers a rare view of early results. The 2015 analysis RIPE NCC Membership: Developments After Reaching the Last /8 reported that 5,191 final-/22 allocations had been made between 14 September 2012 and 1 January 2015. Because returned blocks could be reallocated, those awards involved 5,175 unique blocks.

The broad pattern supported the rule. Most LIRs held one final allocation. That indicates the account-based cap usually operated as intended at the account level. It also means thousands of LIRs received a small block that could otherwise have been consumed by fewer large awards.

The tail revealed the weakness. The analysis found that 3% of LIRs with final-/8 space had obtained more than one such allocation through transfers, mergers or acquisitions. The largest holder had 21 after opening, merging and closing several new LIR accounts in the final quarter of 2014. In the same quarter, 468 new LIRs joined, 38 had already closed by the start of 2015, and 31 of those 38 were Russian LIRs created by three legal organisations.

Those figures do not show that all multiple accounts were abusive. A corporate group can have legitimate reasons for separate LIRs: distinct businesses, jurisdictions, networks, acquisitions or operational teams. Nor does early closure prove the company never used the addresses. But the pattern demonstrates that an account was not a reliable proxy for a new entrant.

The response also reveals the economics. Proposal 2015-01 introduced a 24-month holding period before recently received allocations could be transferred. Delaying transfer made short-lived account harvesting more expensive because the account and its fees had to be maintained longer. It did not make extra accounts impossible. It changed the carrying cost.

This is a crucial point. Once the rule attaches a valuable /22 to an account, the membership fee becomes part of the acquisition price. The registry may not sell addresses explicitly, but an organisation can compare the cost of opening and maintaining an account with the market cost of 1,024 addresses. If the former is lower, policy creates an arbitrage. Holding periods, merger rules and closure treatment then become anti-arbitrage instruments.

The last-/8 policy did not simply ration addresses. It transformed membership architecture into a market variable.

Membership growth was not a clean entrant denominator

By 2019, the distinction between accounts and entrants was impossible to ignore. RIPE NCC's Financial Report 2019 said the number of LIR accounts rose from 20,624 to 25,125 during the year, an increase of 4,501. It explicitly identified impending IPv4 exhaustion and the coming waiting-list regime as strong triggers.

The same report said that, from early 2016 onward, approximately 30% of new LIR accounts were additional accounts. That single figure prevents a celebratory reading of membership growth as new-network growth. Some new accounts represented established organisations buying another chance at the ration.

The 2019 Annual Report adds detail. RIPE NCC made 6,197 /22 allocations in 2019, and 5,885, nearly 95%, went to new LIR accounts. Again, "new account" is not "new firm." The figures show a rush to secure the remaining entitlement before exhaustion. They do not establish how many independent networks entered, how many were controlled by existing members, or how many /22s supported new services rather than later consolidation.

This does not make the statistics useless. They show that the rule created strong demand and that the final pool continued serving accounts until its last year. They also show why the proper denominator must be richer. RIPE NCC should have reported allocations by independent legal organisation, corporate group where verifiable, new versus existing network, additional account, later merger, closure, transfer and operational use.

Privacy and corporate complexity limit perfect classification. The answer is not to publish confidential ownership files. Aggregate cohorts would be enough: first account for an organisation, additional account for an existing member, related legal entity, recent incorporation, merger successor and unresolved classification. The categories could be audited without naming every company.

Outcome measures should follow the cohort. Did the recipient announce the block? Did it also originate IPv6? Did it transfer or merge the allocation after the holding period? Did the account remain open? Did the network obtain further IPv4 on the market? Did the allocation support a new ASN or merely add inventory to an established group?

Without this evidence, membership growth is a measure of demand for the entitlement, not proof that the policy preserved independent entry.

A fee can become the shadow price of a ration

RIR membership fees are charged for an institutional relationship and services, not as a declared purchase price for addresses. Yet incentives are determined by marginal benefit. If opening an extra LIR gives an organisation access to a /22 worth more than the sign-up and carrying cost, the fee operates as a shadow price for the entitlement.

The 2019 figures illustrate the effect without requiring a speculative market price. RIPE NCC recorded unusually strong membership income and expected consolidation after run-out. It also expected the revenue associated with additional accounts to be sensitive to that consolidation. The institution's financial position had become linked to a temporary scarcity strategy used by members.

This creates a governance conflict even without misconduct. The registry wants a sustainable member base and predictable revenue. The final-/8 rule encourages accounts that may disappear once their allocations can be consolidated. Policy restrictions intended to preserve the pool also extend the period during which additional fees are paid. A holding period can protect a legitimate anti-speculation objective while producing revenue for the institution that enforces it.

The answer is not to infer bad faith. It is to separate the accounts. RIPE NCC should report how much income came from first LIRs and how much from additional LIRs opened after the final-/8 trigger. It should show expected closure and consolidation, the cost of processing those changes, and any surplus attributable to address-driven growth. Members can then judge whether a charging scheme amplified the incentive or merely recovered service cost.

Fee design could also have reduced distortion. One option would have been a corporate-group cap, but that would require difficult ownership judgments. Another would have charged additional accounts more nearly in line with the cost and scarcity effect, though this risks making wealth the explicit allocation rule. A third would have used a non-transferable transition entitlement attached to a verified organisation, but later corporate change would become hard to administer. Every alternative carries governance cost.

The important lesson is that "one per LIR" was not neutral simply because LIR was an existing administrative category. It turned entity design, account fees and time into substitutes for a direct market price. Sophisticated holders could optimise those variables. A genuine new entrant usually could not.

Rationing does not remove price. It changes where price appears.

The 24-month hold slowed arbitrage, not exclusion

The accepted 2015-01 policy aligned transfer requirements by imposing a 24-month holding period on newly received allocations. The measure addressed a visible strategy: open an LIR, obtain the /22, move it and close the account.

The hold increased cost in three ways. The account had to remain open longer. Capital tied to the anticipated transfer remained less liquid. Organisational restructuring could not immediately consolidate the allocation without encountering the restriction. These effects made the shortest harvesting strategy less attractive.

They did not make the /22 more adequate for a legitimate entrant. A new network that needed more than 1,024 addresses still had to find another source. An existing group willing to carry additional accounts for two years could still compare that cost with the transfer market. The restriction changed timing and liquidity, not the underlying difference in historical holdings.

It also risked burdening good-faith events. A start-up can be acquired within two years. A group can reorganise. A network plan can fail. An operator can discover that it needs a different structure. Anti-arbitrage rules should distinguish sham cycling from genuine corporate change, but evidence of intent is difficult and discretionary review can itself create risk.

The later 2016-03 "Locking Down" proposal shows the pressure. It proposed a special final-allocation status and restrictions including transfer limits. Its own analysis said organisations using additional accounts would have to keep paying annual fees if they could not consolidate. The proposal was withdrawn in November 2016. It remains useful evidence of the available policy choices and the absence of agreement on a harder lock.

The withdrawn status matters. The proposed restrictions must not be described as adopted law. They reveal an institutional dilemma, not a final rule. RIPE could protect the ration by making the block less portable, but reduced portability would also make the asset less useful and tie holders more closely to account structures. It could permit transfer, but then the ration could be harvested and consolidated. Scarcity turned every preservation device into an allocation of cost.

The 24-month hold slowed one route around the cap. It could not preserve open entry after the pool ended, and it could not equalise the entrant's position while the incumbent market continued.

The transfer market was not outside the last-/8 system

RIPE NCC's 2019 timeline records the first in-region IPv4 transfer on 17 October 2012, barely a month after the final-/8 regime began. Rationing and transfer were not successive eras. They developed together.

That coexistence makes economic sense. The /22 set a maximum on the low-cost primary channel. Any organisation needing more had to reduce demand or use another source. Transfers allowed previously allocated space to move toward buyers. Leasing, provider assignments, acquisitions, sharing and IPv6 supplied other adjustments. The final-/8 rule and the market were one system of entry.

Early transfer evidence shows rapid growth. A 2015 RIPE Labs analysis reported 2,252 unique blocks transferred between October 2012 and May 2015. By 2019, the RIPE NCC Annual Report recorded 3,034 blocks transferred within the service region that year: 2,339 provider-aggregatable blocks comprising 7,844,864 addresses and 695 provider-independent blocks comprising 573,056 addresses. It also recorded inter-RIR flows.

These numbers are not directly comparable to /22 allocation counts. A transferred block can have any permitted size, and one transaction can involve multiple blocks. Some transfers reflect corporate restructuring rather than an arm's-length sale. Public records do not include contract price. The figures nevertheless show that significant quantities moved outside the final-pool ration.

Later empirical research found that transferred space was generally routed and utilisation tended to rise after transfer. This does not prove every market outcome fair or efficient. It challenges the idea that the market was merely hoarding. Buyers acquired addresses because the primary channel no longer met demand.

For the entrant, transfer introduced capital and diligence. It had to find a seller, assess control, inspect reputation, agree price, prepare evidence, manage payment risk and wait for recognised transfer. The incumbent with excess holdings had an asset. The entrant had a financing requirement. The /22 softened the first step but did not alter that asymmetry.

Transfer recognition therefore became RIPE NCC's pro-entry duty. The registry could not supply more space, but it could make reassignment accurate, predictable and timely. A clear transfer path lowers search and legal cost. An opaque or discretionary path compounds the scarcity premium. The institution's role should be to protect the record, not to judge whether the buyer's planned growth deserves capital.

The rule did not fail because a market emerged. The market was how unmet demand continued to find supply. The accountability issue is whether the ration and the market were evaluated together.

Incumbent holdings set the real expansion condition

The final /8 contained a small fraction of the IPv4 space already allocated across the RIPE NCC service region. That historical stock determined the expansion market.

An incumbent with ample holdings could use addresses internally without paying the current transfer price. Economically, the addresses still had an opportunity cost because they could be sold or deployed elsewhere. But the operator did not face the entrant's cash closing, transfer uncertainty or immediate broker expense. It could sequence conservation and IPv6 deployment around an established customer base.

The entrant faced the market at the margin. After the /22, every additional block required a choice. Buy and carry the asset. Lease and accept counterparty and renewal risk. Use provider space and accept renumbering or dependency. Share addresses more aggressively and absorb operational complexity. Acquire a company with holdings and assume its liabilities. Narrow the service. Delay growth.

This difference can affect product design. A hosting provider may ration dedicated IPv4 or charge customers separately. An access provider may deploy carrier-grade translation earlier. An enterprise network may keep services behind shared gateways. A small operator may reject a customer whose application cannot tolerate address sharing. These are market effects of scarcity, not direct registry orders, but policy determines how much initial relief and transfer certainty the entrant receives.

Historical holders also influenced supply timing. A block enters the market when its holder expects the sale value to exceed continued use, option value, transition cost and risk. The registry cannot command that release without assuming a much broader mandate. A transfer-friendly record can reduce friction, while forced reclamation can create litigation and operational disruption.

The last-/8 ration therefore could slow exclusion from the primary channel but could not control the market's supply curve. It could distribute 1,024 addresses to a new account. It could not compel an incumbent to sell the next 4,096 at an affordable price. It could not erase the wealth embedded in early allocations.

The strongest policy claim should have been modest: preserve a minimal technical foothold while building a transparent and low-friction transfer environment. Claims of fairness beyond that need evidence the available reports do not provide.

IPv6 was a condition, but not a measured cure

Proposal 2010-02 required an LIR seeking its final /22 to have already received an IPv6 allocation from RIPE NCC or an upstream LIR. The logic was clear. The ration was meant to support transition rather than become a substitute for it.

Holding an IPv6 allocation is evidence of an administrative step. It is not proof that the network has deployed IPv6 to customers, carries meaningful traffic, trains support staff, updates security controls or can operate without IPv4. RIPE NCC's early membership analysis asked whether recipients had advanced IPv6 deployment and found a more complicated picture than a simple policy success.

This does not make the condition pointless. It ensured that a recipient had at least engaged with IPv6 addressing and could not claim complete ignorance of the alternative. It also linked the scarcity benefit to a stated transition objective. The requirement was low-cost compared with a detailed deployment audit.

The limit is attribution. A network may deploy IPv6 because of customer demand, procurement, platform support, government requirements, engineering preference or expected IPv4 cost. It may hold an allocation without using it. A correlation between a final /22 and IPv6 resources does not prove the ration caused production deployment.

A better evaluation would follow cohorts. At allocation, record whether the LIR had an IPv6 allocation, originated an IPv6 prefix and offered customer reachability. Recheck after one, two and five years. Compare first-time LIRs, additional accounts and existing networks. Measure IPv6 traffic where voluntarily available, not only route presence. Keep commercial confidentiality intact.

The entrant still needs IPv4 during a dual-stack world because other parties decide compatibility. It cannot unilaterally make every customer, vendor and remote service reachable over IPv6. The cost of IPv4 therefore persists even when the entrant deploys IPv6 well. Treating IPv6 as proof that 1,024 IPv4 addresses are sufficient shifts the external compatibility burden onto the new network.

The condition pointed in the right technical direction. It did not transform the /22 into a scalable entry asset.

November 2019 delivered the policy's clearest result and limit

RIPE NCC's run-out chronology shows the final stages. On 2 October 2019, contiguous /22s were no longer available, so equivalent quantities were assembled from smaller prefixes. On 25 November, RIPE NCC made the final /22-equivalent allocation from its available pool and activated a waiting list for recovered space.

The exhaustion announcement said thousands of new networks had received /22 allocations since 2012. That is the strongest case for success. The cap extended a primary allocation opportunity across seven years and prevented earlier large applicants from consuming 185/8 immediately.

The same announcement acknowledged the limit. The waiting list would provide a single /24, or 256 addresses, only to LIRs that had never received an IPv4 allocation from RIPE NCC. Returned amounts would not approach the millions of addresses networks needed. Transfers and carrier-grade translation had already become prominent responses.

The ration became smaller because there was less to ration. The policy could preserve an entitlement only by reducing its quantity and making timing dependent on returns. This is not policy failure in the sense of avoidable waste. It is the arithmetic endpoint of a finite pool.

It is, however, the end of the stronger fairness claim. Once entry depends on a waiting-list /24 of uncertain timing, the primary channel no longer supplies a broadly scalable input. New operators compete through architecture and capital. Existing holders decide whether and when stock reaches the market. RIPE NCC maintains the recognised record and sets conditions around movement.

The seven-year extension should be valued for what it provided: time, a first block and a visible transition signal. It should not be treated as evidence that exclusion was solved. Exclusion moved from "no allocation today" to "no further low-cost allocation after the ration." The later date mattered, but the economic boundary remained.

The missing study is the entrant cohort, not another pool countdown

RIPE NCC has published valuable counts, timelines and policy records. The missing evidence is a cohort study capable of connecting the /22 to actual entry.

Begin with every final-/8 recipient and classify the account at the date of allocation: first LIR of an independent organisation, additional LIR of an existing member, related company, merger successor, sponsor, uncertain affiliation. Preserve confidential evidence while publishing aggregate totals and classification confidence.

Then measure the address event. Was the allocation a contiguous /22 or an equivalent assembled from smaller ranges? How long did approval take? Was the block announced within six months? Did route origin match the recipient or a provider? Did the account later transfer, merge, close or return the space?

Add the growth path. Did the organisation acquire more IPv4 through recognised transfers? Did it lease or receive provider space where observable? Did it obtain a larger address portfolio through acquisition? How long after the first /22 did it need more? The registry need not demand private business plans; it can use its own recognised events and carefully labelled routing observations.

Add IPv6 without overstating it. Did the organisation originate IPv6 before the /22, shortly after or never? Did the IPv6 route remain visible? Where voluntary operator data exists, did customer traffic grow? A route is evidence of deployment capability, not complete adoption.

Finally, compare outcomes. First-account recipients should be compared with additional accounts and with similar organisations that entered after 2019. Measures could include account survival, transfer acquisition, routing continuity and time to additional IPv4. The study should not pretend that policy caused every difference. It can reveal whether the ration reached the group it was meant to assist.

The counterfactual also matters. What would have happened under /21, /23, a corporate-group cap, an auction, a transferable voucher or immediate exhaustion? Each alternative changes quantity, duration, administrative burden and susceptibility to strategic behaviour. There is no costless design.

Without this study, "thousands of networks" remains an institutional claim based largely on account and allocation counts. It may be directionally true. It is not precise enough to establish competitive entry.

What membership accountability required

RIPE NCC is a membership association as well as a registry. The final-/8 episode tested whether those roles could be kept distinct.

Members had influence over charging and governance, while the open RIPE policy process developed address policy. Yet the people who benefited from extra accounts, held large historical portfolios or needed future entry did not have identical interests. A count of LIRs could be changed by the scarcity rule itself. When one organisation opened several accounts, it increased both the allocation denominator and potentially the institutional membership base.

Accountability required disclosure before policy adjustment. How many allocations went to first-time organisations? How many went to additional accounts? Which member cohorts supported or opposed transfer holds? How much fee income came from address-driven accounts? What operational cost did the surge impose? How many accounts consolidated when the hold expired?

It also required a conflict statement. RIPE NCC staff had legitimate expertise and had to explain implementation effects. The institution also received revenue from the accounts and administered the transfer restrictions. Its analysis should therefore separate facts, forecasts, recommendations and financial incidence.

Decision rights needed clarity. The policy forum could define distribution from the final pool, but changes affecting membership contracts, fees, mergers and transferability crossed institutional boundaries. Members should know which body could change each condition, what participation supported the change and how an affected holder could seek review.

The legitimacy test is not whether the policy was discussed openly. It is whether the rule's unit matched its stated beneficiary, whether effects were measured and whether the institution disclosed its own exposure. A public mailing list cannot turn an LIR account into an independent entrant by assertion.

The 2012 policy was administratively elegant because it used an existing unit. Membership accountability required admitting when that unit stopped representing the social claim attached to it.

A narrow registry can be more pro-entry than a generous gatekeeper

Once the free pool is gone, the registry's most valuable pro-entry service is not rationing. It is reliable recognition.

A buyer needs a clear evidence list, predictable review, fraud protection, reasoned requests for more proof and an effective appeal. A lessee needs operational contacts and route-security arrangements that reflect actual roles. A new member needs to understand the difference between membership, sponsorship, provider space, transfer and waiting-list eligibility. A company undergoing acquisition needs continuity while control records change.

These are narrow functions with large economic effects. They reduce the fixed cost of entry without choosing which company deserves to win. They make old holdings more liquid, which can increase supply. They protect sellers from impersonation and buyers from duplicate claims. They help lenders and investors distinguish operational risk from administrative uncertainty.

A generous gatekeeper offers a small ration but retains broad discretion over later movement. A narrow registry may offer no free stock yet provide fast, accurate and reviewable settlement. In a mature scarcity market, the second can be more useful.

This does not mean automatic approval. Contested authority, sanctions, court orders, fraud signals and inconsistent records require judgment. The decision should stay connected to the record. The registry should not assess whether the buyer's service is socially worthy, whether leasing is morally acceptable or whether regional scarcity should override a valid transfer.

Service performance should expose difficult cases. Publish transfer times by percentile, requests for additional evidence, withdrawals, refusals, appeals and reversals. Separate routine transfer, merger, insolvency and dispute. Report inter-RIR incompatibility. Averages conceal the cases in which a new entrant's financing is most exposed.

The last-/8 rule tried to preserve entry by keeping supply in the institution's hands. The post-pool task is to preserve entry by keeping the institution's hands disciplined.

NRS should campaign for the exit that the /22 could not provide

The Number Resource Society cannot guarantee an entrant cheap IPv4 or provide an administrative exit. It is a global member-representation and advocacy organisation, not a registry, RIR, transfer authority or continuity operator. Its legitimate role is to document the consequences of scarcity, convene affected operators, publish evidence-led comparisons and campaign for safeguards that the authorised institutions must enact.

The first safeguard NRS can advocate is portability of the verified record. If a registry becomes unreliable, insolvent or captured, an RIR governance process, court or other legally empowered authority should provide a route by which the holder can carry signed history, identity evidence and uncontested status to a qualified successor. A disputed claim remains labelled and protected; portability is not permission to erase conflict, and NRS cannot decide the claim or select the successor.

The second safeguard is transfer recognition limited to necessary facts. The responsible RIR verifies the transferor, transferee, range, authority, conflict state and effective change under the applicable policy. Commercial price and business purpose stay outside unless law specifically requires otherwise. The registry records the recognised change; an advocacy group neither settles the transaction nor allocates capital.

The third safeguard is continuity. A disagreement over membership fees, policy interpretation or documents should not casually disable routes, reverse DNS or RPKI services on which customers rely. The RIR or other authorised service operator should scope and time-limit holds, with review by the competent independent body. Fraud emergencies require fast containment by the operator that controls the service, followed by review; NRS can support affected members and scrutinise the published process, but it cannot operate the controls.

The fourth safeguard is transparent cost. Operators should pay the authorised provider for verification, publication, security, correction, interoperability and recovery. They should not pay a scarcity rent proportional to the value the registry can obstruct. Additional services require lawful governance approval, while NRS advocacy is separately funded and does not become a registry charge merely because NRS campaigns on the issue.

The fifth safeguard is an accurate denominator. In its own research and advocacy, NRS should distinguish a network, legal organisation, account, mandate and resource holding. One organisation may legitimately have several of each, but public claims must not count them as interchangeable evidence of representation or entry.

This advocacy would not have created more addresses in 2012. It could have made the limits of the /22 clearer and built member pressure for a safer route beyond it, while implementation remained with RIPE NCC and any other competent authority. The positive policy objective is not abundance. It is that a new operator can buy, lease, transfer, secure and prove control through narrow, reviewable registry procedures rather than through powers exercised by NRS.

The verdict: useful delay, unequal entry

RIPE NCC's final-/8 rule deserves neither a triumphal verdict nor a cynical dismissal.

It succeeded at a narrow and important task. A standard cap prevented a small number of large justified requests from consuming 185/8. Thousands of LIR accounts received a routable /22 over seven years. The rule created time for networks to deploy IPv6, sharing and transfer strategies. It made the final phase predictable.

It failed as a complete answer to entry because no ration could erase historical distribution. A /22 was a foothold, not a growth portfolio. The account-based unit could be multiplied. Holding periods raised carrying cost but did not eliminate strategic structure. Membership growth included additional accounts. Incumbents held the supply from which entrants had to obtain their next addresses. Transfer capital and administrative recognition set the real marginal conditions.

The policy's deepest lesson is that rationing cannot remain morally self-sufficient after scarcity creates a market. Keeping the last pool alive is not the same as keeping competition open. The first is measured in allocation dates and prefix counts. The second is measured in independent entrants, financing cost, transfer access, time to scale and the burden of substitutes.

RIPE NCC should be credited for publishing evidence that exposes the ambiguity: multiple accounts, 21-block concentration, additional-account shares, transfer volumes and post-run-out consolidation. Those records allow a more honest conclusion than the institutional slogan.

The new entrant was not saved from scarcity. It was given a small piece of time. What happened next depended on capital, architecture, incumbent supply and the reliability of the ledger. The correct reform is not a different moral ration from an empty pool. It is a market-compatible registry with narrow authority, measurable service, real appeal and an exit path.

That would preserve the thing the /22 could not: not equal quantities, but an equal right to move, prove and continue operating under scarcity.

Sources