Summary
- LACNIC moved through four operational phases: Phase 0 began in October 2013, Phase 1 on 19 May 2014, Phase 2 on 10 June 2014 and Phase 3 on 15 February 2017. Each transition changed more than block size. It changed who could apply, how often, what happened to incomplete requests and what queue position was worth.
- On 19 August 2020, LACNIC reserved the last immediately available Phase 3 block. Its 2020 annual report records a doubling of average assignments in the first half of August, a record 234 new members that month, 307 IPv4 requests in the third quarter and creation of a waitlist fed only by returned or recovered addresses after a 180-day quarantine.
- LACNIC reported in October 2020 that roughly 600 requests were pending, about 200 already on the waitlist and 400 still in process. It expected only 170 organizations to receive space during the next twelve months and estimated recoveries at five or six /22-equivalent blocks a month. Those figures reveal a demand funnel, not one simple line.
- Queue visibility was useful but incomplete. A public position could show order among admitted requests, while leaving outsiders unable to reconstruct incomplete, withdrawn, rejected, merged, unpaid or never-filed demand. Brazil and Mexico also used their own national registry channels, so no single LACNIC list represented the entire region.
- Entry cost moved from an administrative fee plus documentation to a package of membership, IPv6 qualification, uncertain waiting time, financing delay, address-sharing equipment, secondary-market procurement and the risk that a /22 would be too small by the time it arrived. A nominally free allocation could therefore carry a substantial economic price.
- Accountability after exhaustion requires more than publishing the head of a line. LACNIC should preserve an anonymized event history for every cohort, including admission, completeness, approval, deferral, block size, payment lapse, withdrawal, rejection reason and release source, while reporting how many qualified organizations remain unserved and how waiting time is distributed.
The last block did not end allocation; it changed what was being allocated
The date 19 August 2020 is easy to narrate as a finish line. LACNIC assigned or reserved its last immediately available IPv4 block, the regional free pool reached zero, and a long-forecast technical event finally arrived. That account is correct as inventory history. As institutional history, it stops too soon.
Before the last block, an approved applicant was competing for a stock that could ordinarily be drawn down. After it, an applicant was competing for a place in time: a conditional position against future returns and recoveries whose size, timing and usability were unknown. The scarce item was no longer only an IPv4 prefix. It was priority over an uncertain future release.
That distinction matters because a registry can publish the amount it allocated while failing to show the opportunity it denied. Allocations are observable completions. Scarcity is also made of delayed networks, scaled-back launches, carrier-grade address sharing, provider dependence, secondary-market purchases and projects that never entered the formal line because the expected wait was too long. A completed-allocation chart contains none of those counterfactuals.
LACNIC's own exhaustion page describes Phase 3 as a pool composed of post-exhaustion IANA allocations and recovered or returned blocks. Once the ordinary pool was exhausted, only those dynamic sources remained, alongside a reserve for critical infrastructure. Since March 2020, recovered resources have had to spend six months in quarantine before gradual release. The same page says a member must not previously have received IPv4 space, must first receive IPv6 resources, and can join the waitlist without any guarantee of ultimately receiving IPv4.
Those are not minor conditions attached to a conventional order. They define the product. The applicant receives neither a fixed delivery date nor a certain quantity beyond the permitted /24-to-/22 range. It receives an opportunity to be assessed when suitable recovered stock approaches release. A queue position is therefore closer to a contingent option than to an invoice awaiting fulfilment.
The policy had defensible aims. It preserved some possibility of IPv4 access for organizations arriving late, made IPv6 part of the entry path, limited the rate at which the final reserve could disappear and prevented one large approved request from absorbing the remainder. But defensible aims do not answer the distribution question. Which organizations bore the delay, which obtained the last predictable stock, which disappeared before service, and which costs moved outside the registry's accounts?
That is the accountability problem concealed by the phrase "the pool was exhausted." The pool had a visible ending. Demand did not.
Four phases created four different meanings of first come, first served
LACNIC did not move from abundance to a waitlist in one step. Its published phase history records Phase 0 beginning in October 2013, Phase 1 on 19 May 2014, Phase 2 on 10 June 2014 and Phase 3 on 15 February 2017. The labels imply a smooth countdown. The rules show a sequence of different allocation institutions.
Phase 0 covered the period until the inventory reached the last available /9. Requests were handled through tickets on a first-come, first-served basis, with NIC Mexico and NIC.br managing their tickets independently. Need remained the substantive standard. Large requests received joint review: applications for a /15 or more involved the national registries and LACNIC, while a /16 equivalent was assessed by two hostmasters within the same organization. An incomplete request went back in the line after the applicant supplied the missing information.
Phase 1 began once the remaining inventory approached the /10 protected for gradual exhaustion. Need still governed approval, but aggregation could no longer be promised. An organization qualifying for a /14 might receive several prefixes with the same total capacity. That fragmentation imposed a technical and commercial cost even when the number of addresses matched the approved need: more route announcements, more records, more operational handling and potentially more filtering or reputation work.
Phase 2 activated the /10 reserved for gradual exhaustion. The maximum shrank to a /22 and the minimum to a /24. Existing and new organizations could request additional space, but only after six months. A business needing a /19 no longer presented one need and received one answer. It faced a sequence of small decisions spread across time, each exposed to continued availability and renewed review.
Phase 3 changed eligibility itself. The final reserve was for organizations that had not previously received LACNIC IPv4 space, and the assignment would be their first and last from that pool. The 15 February 2017 announcement put the available range at /24 to /22 and reported 4,698,112 addresses in the Phase 3 reserve at the start, with possible additions from recovery and post-exhaustion IANA distributions.
The same words, first come and first served, therefore did not describe one stable right. In Phase 0, the line ordered applications against a comparatively broad pool under ordinary need. In Phase 1, the line also allocated fragmentation risk. In Phase 2, it ordered small recurring rations and placed a six-month interval between requests. In Phase 3, it excluded prior recipients and ordered first-time entrants against a terminal reserve. After August 2020, it ordered conditional access to future recovered stock.
Changing rules at transparent thresholds can be fairer than improvisation. Yet a phase transition creates winners and losers at the boundary. Two otherwise similar organizations can face radically different possibilities because one ticket became complete before a threshold and the other after it. A large incumbent could receive meaningful space under Phase 1, return six months later for a /22 under Phase 2, and then become ineligible under Phase 3. A new entrant arriving after the final free block could satisfy every substantive condition and still wait for years.
The phase label is thus an allocation fact with economic force. It determines the size of the ration, the frequency of access, the value of prior membership and the point at which waiting replaces delivery.
A timestamp was only one of several clocks
Queue discussions often assume that the earliest timestamp wins. LACNIC's procedures were more layered. The ticket arrival time mattered, but so did completeness, pre-approval, hostmaster review, payment, the signed service agreement, block availability and the applicable phase at the decisive moment.
In Phases 0, 1 and 2, incomplete applications were sent to the end of the relevant ticket line. That rule has an intuitive justification: an applicant should not freeze scarce capacity while supplying missing evidence at leisure. It also converts information quality into priority. An operator with experienced consultants, established corporate documents and staff accustomed to registry requests can preserve an earlier effective position more easily than a small new network encountering the process for the first time.
Phase 2 added a pre-approval layer. Once requirements were met, hostmasters entered request information into a shared form. The request did not become final until payment and a signed agreement arrived. If payment was not received within fourteen days, the prefix lost its reserved status, returned to the central pool and could be pre-approved for another organization. At the end of the /10, a committee representing LACNIC and the two national registries was to review the UTC order and requested amounts among pre-approved cases to determine which could be satisfied.
This reveals at least five clocks.
The first clock begins when a ticket is received. The second measures how long the applicant takes to make it complete. The third measures registry analysis and any repeated information exchange. The fourth begins at pre-approval and ends with payment and contract completion. The fifth is the inventory clock, which can push an otherwise qualified case across a phase boundary.
The clocks do not burden every applicant equally. A financing committee may not release funds within fourteen days. A cross-border company may need several signatures. A local provider may need to translate or formalize records. A founder may be building the network while answering technical questions. These are not reasons to waive verification, but they are reasons to report attrition. If unpaid reservations, incomplete tickets and withdrawn cases vanish from the public account, the institution appears to have met every demand it chose to count.
After August 2020, another clock became decisive: quarantine. Recovered addresses could not immediately satisfy the next request. They waited 180 days, allowing operational references, routing, reputation and prior use to cool before reassignment. That protection has value. It also means the date of recovery is not the date of supply. Applicants are exposed to both the uncertain arrival of a block and a fixed delay before it can re-enter circulation.
A public line ordered only by admission date cannot explain these intervals. To understand whether priority was administered consistently, the public needs event times: received, complete, approved, eligible for stock, offered, accepted, paid, assigned, lapsed, withdrawn or rejected. Names can remain confidential. The sequence cannot.
Phase 2 rationed expansion; Phase 3 rationed entry
The economic distinction between Phase 2 and Phase 3 is more important than the shared /22 maximum suggests.
During Phase 2, the policy allowed an organization to request further addresses after six months. The /22 ceiling constrained each allocation, but an existing member could remain in the allocation channel if it continued to qualify. This rationed expansion. It slowed the rate at which a growing operator could obtain addresses from the registry and encouraged it to economize, share addresses, acquire space elsewhere or stagger growth.
During Phase 3, prior LACNIC receipt became disqualifying. The reserve was no longer a small recurring supplement for established networks. It was a one-time bridge for a first recipient. This rationed entry. The decision favoured breadth of initial access over depth of support for existing operators.
That trade-off deserves to be stated plainly. Giving one /22 to a new member may help it become independently numbered, multihome, run infrastructure or begin a transition strategy. Giving the same /22 to an established operator may connect more immediate customers because the operator already has facilities, demand and staff. The first option protects formal entry. The second may produce faster short-run utilization. Neither priority follows automatically from packet routing.
The new-entrant reserve also changed the value of organizational history. Having received IPv4 earlier became a liability for access to the final pool. Not having received it became an eligibility asset. Such a rule can create pressure around corporate identity, mergers, acquisitions and related entities. LACNIC's current waitlist notice addresses one part of this problem by making a position non-transferable and stating that it is lost through sales, purchases or mergers.
Non-transferability reduces a direct market in queue slots. It does not remove the underlying option value. A business can still decide when to create an eligible entity, when to seek membership, whether to remain independent through the waiting period and how to structure upstream service while it waits. Where one-time access has market value, identity rules become economic rules.
The requirement to receive IPv6 before an IPv4 request enters the waitlist is similarly double-edged. It aligns entry with the protocol that can support long-run growth and prevents a member from joining solely for a speculative IPv4 claim without taking an IPv6 resource. But an IPv6 allocation is not the same as effective IPv6 service to customers. The rule proves an administrative step, not deployment quality, traffic share, application reachability or reduced IPv4 dependence.
These limitations do not make the Phase 3 design irrational. They show why its success cannot be measured only by the number of first-time recipients. The relevant outcomes are whether recipients built durable networks, whether the /22 remained operationally useful, whether related-company arbitrage occurred, how long genuine entrants waited and what happened to qualified applicants who never reached assignment.
The August surge was a rational response to a known cliff
LACNIC's 2020 annual report supplies unusually important evidence about the final weeks. Average assignments doubled during the first half of August. The organization registered 234 new members that month, a record. The third quarter brought more than 307 IPv4 requests, 75 percent more than the previous quarter. The last available block was assigned or reserved on 19 August.
It would be easy to describe this as panic or opportunism. That would be an institutional conclusion without sufficient evidence. When a valuable input is available under known rules and a hard cutoff is approaching, earlier application is rational. A company that expects to need addresses next year has reason to advance membership and documentation this year. An operator that can choose between a predictable /22 now and an uncertain recovered block later has reason to choose the former.
The surge was therefore partly produced by the policy's own time structure. Announcing the cliff improves notice but concentrates demand before it. Restricting Phase 3 to first-time recipients encourages organizations considering independent numbering to establish eligibility before supply disappears. A queue is not merely observing demand; its expected closing date changes demand's timing.
That feedback weakens simple exhaustion forecasts. An estimate based on average prior assignments assumes applicants will behave similarly as the end approaches. They will not. The closer the cutoff, the more valuable an early position becomes. An expected last block can pull future requests into the present, accelerating the date it seeks to predict.
The official record acknowledges this practical effect without resolving its distribution. The report says LACNIC worked with the national registries on requests in the weeks before and after exhaustion and created a waitlist intended to add transparency by showing ticket receipt and prefix approval dates. What it does not show in the report is the composition of the surge: how many applicants had pre-existing plans, how many accelerated entry, how many became complete, how many were rejected, how many obtained smaller prefixes than they sought and how many joined without receiving IPv4.
Those missing figures matter to membership accountability. The record of 234 new members is institutional growth. It is not automatically evidence that 234 new networks received useful IPv4 capacity. The 2,596 total resource requests processed in 2020 included number requests, returns and transfers. It cannot serve as an IPv4 demand denominator without separation.
The point is not to question every August application. It is to recognize that a deadline converts information about policy into private option value. Applicants able to monitor the pool, prepare documents and finance membership quickly can act on that information. Operators that learn late, lack specialist help or require slower approval enter the less certain regime. Transparency about the cutoff is necessary. Transparency about who could respond to it is the next step.
Six hundred pending requests were not one queue
In October 2020, LACNIC's account of the new exhaustion conditions reported approximately 600 pending IPv4 requests. It said only 170 organizations were expected to receive space during the next twelve months, while the remaining 400 and future applicants would wait longer. The Spanish version gave a more revealing split: roughly 200 requests were on the waitlist and another 400 were still being processed.
That split prevents a common analytical error. The 600 were not six hundred identical, approved claims ordered from first to last. Some had reached the waitlist. Others remained under analysis. A request in process might become approved, return for more information, be rejected, be withdrawn or cease to qualify. The visible waitlist was one segment of a larger demand funnel.
Even within the admitted line, supply was uneven. LACNIC expected to recover only the equivalent of five or six /22 blocks per month, and releases depended on the size and timing of returned space. A recovered /20 can be divided differently from several non-contiguous /24s. Reputation history and technical cleanup can affect usability even after quarantine. The first applicant's justified amount may not fit the next available prefix cleanly.
The current waitlist page now warns that, based on historical recovery behaviour, the last request may wait at least eighteen years and receive at most 1,024 addresses. It also says the estimate is inherently uncertain because future recoveries cannot be known. This is a powerful disclosure. It tells a late applicant that the line may outlast its business plan, equipment cycle, financing and perhaps the relevance of the requested quantity.
Yet the estimate also demonstrates why position alone is a weak accountability measure. An eighteen-year tail can shrink because applicants withdraw, merge, fail, acquire space elsewhere, cease membership or no longer meet need. If the line advances partly through attrition rather than supply, a shorter observed wait does not mean more addresses became available. It may mean demand disappeared.
The public should therefore see cohort survival. For applicants entering in each quarter, report how many were admitted, remained active after one year, received space, withdrew, became ineligible, failed payment, were rejected or left for an identified alternative where voluntarily disclosed. Report the number of addresses requested as well as the number of tickets, because one /24 request and one /22 request are not the same unmet demand.
There is also a regional perimeter issue. NIC.br and NIC Mexico administer requests in their national channels. Their queues and timing do not automatically become one global LACNIC order. This may reflect legitimate operational delegation and local service. It means analysts should not present the central list as the full Latin American and Caribbean demand curve. Comparable national data and a reconciled regional total are needed.
The queue did not literally disappear. The denominator around it did. Published positions made one part visible while the unadmitted, unresolved and exited portions remained difficult to measure.
A free-pool address carried a price even when no purchase price appeared
An address allocated from the remaining pool was not sold at a secondary-market price. That does not make access costless.
An entrant first faced membership and application costs. It needed technical plans, corporate records, staff time, IPv6 resources and the ability to answer questions without losing effective priority. It then faced the time value of waiting. Capital committed to a network could not produce the intended service if public IPv4 capacity was missing. Equipment might sit underused. Customer acquisition might be postponed. Upstream providers could supply addresses, but that increased dependency and future renumbering cost.
Next came substitution. Carrier-grade address sharing can stretch scarce public addresses, but requires gateways, logging, port management, abuse response and support for applications that behave poorly behind shared translation. Buying or leasing addresses elsewhere adds cash cost, diligence, routing reputation risk and contract exposure. Delaying the project sacrifices revenue or service coverage. Each substitute is a price paid because the registry allocation did not arrive.
Market evidence helps establish the scale without pretending that every queued /22 had the same sale value. An APNIC analysis of reported market data found prices around USD 22 per address in 2019 and the first half of 2020, followed by a sharp rise during 2021. At USD 22, a /22's 1,024 addresses would correspond to roughly USD 22,528 before transaction and deployment costs. That is a market benchmark, not a valuation of LACNIC's reserve and not evidence that a recipient could immediately sell its allocation.
The benchmark nevertheless clarifies the queue's option value. Earlier access to a /22 could avoid or defer a five-figure purchase. It could support customers, preserve an upstream negotiation position or create strategic flexibility. A policy that grants such access is allocating an economic advantage even if the invoice calls only for membership and service.
The advantage was not equal for all recipients. A /22 may be transformational for a small provider and immaterial for a mobile operator. A clean contiguous block may be more useful than fragmented supply. An address with adverse prior reputation may cost more to rehabilitate. An allocation received after a five-year delay may be worth less to the original project than one delivered within months.
This is why reporting only fees understates entry cost. The relevant account includes money, delay, uncertainty, substitute technology and foregone scale. It also includes the distribution between incumbents and newcomers. Existing holders entered the scarcity era with an installed base whose acquisition cost had largely been incurred under earlier conditions. A late entrant faced contemporary market value and a queue simultaneously.
The queue was intended to protect new entry, but it could not erase the first-mover advantage created by historical allocation. At best, it provided a small bridge. Accountability begins by refusing to call the bridge equivalent to the road incumbents already possessed.
The allocation record shows supply; the missing-demand ledger shows policy impact
LACNIC's 2019 annual report recorded 5,692 /24-equivalents assigned that year through 1,556 IPv4 assignments. It also reported IPv6 coverage among its members at 94.88 percent. These figures describe a substantial Phase 3 operation before the last free block.
They do not answer how much need remained. An assignment count measures successful cases. A /24-equivalent count measures delivered quantity. Neither shows the requested amount, the share approved at a smaller size, waiting time, failure rate or downstream customers affected.
A useful scarcity account needs four linked views.
The first is the stock account: opening available space, additions, reservations, quarantine, releases, assignments, returns and closing stock by prefix size. This establishes physical conservation and prevents double counting.
The second is the request account: tickets received, applicants, requested addresses, completed cases, pre-approvals, approvals, assignments, rejections, withdrawals and lapses. This establishes the demand funnel.
The third is the time account: median and tail time from receipt to completeness, completeness to decision, approval to assignment and waitlist admission to offer. This shows whether delay comes from the applicant, the registry or absent stock.
The fourth is the outcome account: recipient type, network activation, routing observation with appropriate caveats, IPv6 deployment, subsequent transfer or corporate change, and whether the allocated quantity remained proportionate to the declared bridge need. This tests whether the policy's intended entry benefit appeared in operation.
None of these views requires publishing confidential business plans. Applicant identifiers can be pseudonymous. Requested quantities can be placed in bands. Rejection reasons can use stable categories. Very small cohorts can be combined to prevent identification. The objective is not public re-adjudication of each case. It is evidence that comparable cases met comparable treatment and that the institution can explain the fate of scarce stock.
The report should preserve unsuccessful demand rather than deleting it after closure. If an applicant withdraws because it acquires addresses elsewhere, that remains evidence of scarcity. If it fails to answer, that may reflect application burden and should be classified separately from a substantive denial. If it cannot pay during a short reservation window, that is a financing failure, not proof that no need existed.
This missing-demand account would also improve policy review. Members could see whether an eighteen-year estimate arises from accelerating admissions, slow recovery, large requested quantities, low attrition or administrative bottlenecks. They could test whether the IPv6 prerequisite filters speculative entry or merely adds a formal step. They could compare the central channel with the national registries without assuming operational differences are unfair.
The existing public line was a meaningful transparency measure. It should be treated as the first column of the account, not the whole account.
Fairness cannot be reduced to preserving the earliest ticket
First-come, first-served has a strong appeal because it appears neutral. The institution does not rank applicants by political importance, employment, nationality, sector or persuasive power. It follows time.
But time priority is fair only if access to the clock is reasonably equal and the clock measures the relevant event. A ticket submitted first may be incomplete. A later organization may have urgent public-service need. An applicant may obtain better advice about the exact evidence required. A large corporate group may create a dedicated eligible entity more readily than a community provider. A member in Brazil or Mexico enters a different administrative channel.
Replacing time priority with broad discretion would be worse. A registry should not decide that one commercial model is more virtuous than another or that a favoured sector deserves hidden preference. The answer is to narrow discretion while making the time rule more complete.
Admission criteria should be objective and published. Completeness questions should cite the requirement they implement. If a request loses position, the event and reason should be recorded. Phase transitions should apply through a declared rule for cases already received, already complete and already pre-approved. Applicants should know whether the decisive date is submission, completion, approval or assignment.
Review should be available for procedural error without converting every scarcity loss into a promise of addresses. An independent reviewer can determine that the registry misapplied order or eligibility. If stock has gone, the remedy may be restored priority against the next suitable release, a fee credit or a reasoned declaration. Secretly displacing another qualified applicant would reproduce the original unfairness.
Related entities require special care. A one-time allocation policy needs a beneficial-control test to prevent simple duplication, but the test should not treat every investor, supplier or commercial relationship as common control. Publish the factors: ownership, voting rights, management, consolidation, network operation and contractual authority. Give applicants notice and a chance to correct the record.
Non-transferable queue positions are sensible because a trade in priority would let capital buy the front of a nominally equal line. Corporate changes should still have proportionate treatment. A genuine restructuring that preserves the same operating project is different from selling a place to an unrelated buyer. Automatic cancellation in every merger can punish ordinary financing; automatic succession can create a market in disguised acquisitions. Reasons and review matter.
Fairness is therefore not a choice between a mechanical queue and arbitrary judgment. It is a design in which narrow judgment is bounded, recorded and reviewable, while the ordering rule includes every transition that can materially change priority.
Membership growth should not be mistaken for scarcity relief
The final phases produced a striking increase in membership. LACNIC's 2020 report put the year-end total at 11,657, with 1,604 new members during the year. Its 2021 report recorded 12,144 organizations. These totals can indicate regional growth and wider access to independent number resources.
They can also reflect the institutional packaging of scarcity. If a new organization must become a member and receive IPv6 before joining the IPv4 waitlist, demand for the scarce item raises the number of accounts. Membership becomes both a service relationship and an entry ticket to a contingent claim.
That does not make the members artificial. IPv6 and AS number services have independent value. Many organizations would rationally seek direct registry relations even without IPv4. The analytical point is that membership count cannot prove the exhaustion policy met entrants' IPv4 needs.
Members should see the cohorts separated. How many joined with IPv6 only? How many sought an ASN? How many entered the IPv4 waitlist? How many later received IPv4? How many remained active without it? How many left? How did fees differ while the benefit was pending?
The fee question became visible in 2020. LACNIC reported that once IPv4 was exhausted, fees for new IPv6-only members were significantly higher than before, prompting a phased six-year discount approved late that year. This is a revealing incidence problem. Earlier members often received IPv4 and IPv6 within one relationship. Later members could pay for membership while receiving only the abundant protocol and waiting for the scarce one.
A discount can soften the imbalance. It does not define the correct price of an uncertain place. The registry still provides IPv6 registration, support, certification and other services. Those have cost. The member may also value governance participation and direct records. The fee should therefore be decomposed rather than described as payment for an IPv4 chance.
An accountable invoice would distinguish standing registry services from transaction charges and any waitlist administration. A member should not infer that annual payment preserves a guaranteed IPv4 entitlement if no such entitlement exists. Equally, the institution should not use the growing waitlist to support a broad budget without showing the incremental cost of administering it.
Membership accountability means making the bargain legible. The late entrant pays for real services, receives real rights and may hold a conditional priority. Those three things should not be blurred into one reassuring membership total.
The queue and the secondary market became one entry system
Once recovered supply became too slow for ordinary business timing, applicants did not cease needing IPv4. They moved to substitutes. Some used provider addresses. Some shared public addresses across customers. Some bought or leased space. Some altered launch geography or delayed service. The registry queue and the market therefore formed one entry system even though their records remained separate.
The two channels affected each other. A credible near-term chance of a /22 could reduce an applicant's willingness to pay in the market. An eighteen-year estimate makes market procurement more urgent. Higher prices increase the value of preserving queue eligibility. A slow registry release can support scarcity rent for existing holders, while an active transfer channel can prevent the queue from becoming a complete barrier.
This does not mean LACNIC should price free-pool addresses at market value. Doing so would convert a new-entrant reserve into an auction and favour applicants with capital. Nor does it mean transfers should be ignored as morally suspect. They are evidence that operators value continuity and timing enough to pay for it.
The registry's narrow job is to maintain uniqueness, accurate control records and secure changes. It should verify that the same addresses are not assigned twice, that the requesting party has authority and that the public registration can be relied on. It should not claim that a small allocation eliminates market need or that a market purchase proves an applicant lacked legitimate operational demand.
Policy evaluation should combine the channels without merging their legal character. Report recovered-space assignments, intra-regional transfers, inter-regional transfers, waiting time and known request exits in the same scarcity dashboard. Do not infer a causal explanation for every exit. An applicant can voluntarily state that it acquired space elsewhere, but silence should remain unclassified.
The result would show the true entry boundary. A new operator can obtain independent IPv4 through a recovered allocation, a transfer, a lease or another holder's network. Each route carries different rights, costs and dependencies. The policy's achievement is not that one route exists. It is whether entrants can build without hidden discrimination, indefinite uncertainty or avoidable control by incumbents.
What LACNIC should publish before calling the line transparent
The strongest reform is not a longer explanation of the existing waitlist. It is a small set of reusable records that let members reconstruct each cohort without seeing confidential submissions.
First, publish a daily stock history by source and state: ordinary reserve, critical-infrastructure reserve, returned, recovered, quarantined, approved, reserved, released and assigned. Identify prefix size and aggregation without exposing a future recipient before assignment.
Second, publish a request event table with a random persistent identifier. Record channel, application type, requested size band, receipt, completeness, first decision, waitlist entry, offer, final result and reason code. Keep Brazil, Mexico and the central channel distinguishable and publish reconciled regional totals.
Third, publish the ordering rule in executable prose. State how incomplete cases move, how ties are broken, how available prefix size interacts with justified quantity, how corporate changes affect identity, what happens at phase boundaries and whether a declined smaller block preserves position.
Fourth, publish the unmet-demand denominator. Count active applicants and requested addresses, but also cohort withdrawals, ineligibility, rejected cases and members that close before service. Separate demand rejected on substance from demand that leaves because no timely supply exists.
Fifth, publish waiting-time distributions rather than one estimate. Show median, 75th, 90th and oldest active request, plus completed and exited cohorts. An eighteen-year tail should not be compressed into an average dominated by earlier recipients.
Sixth, publish decision consistency. For each reason code and size band, show the number approved, reduced, returned for information and rejected. Commission periodic independent sampling of whether the stated rule matches recorded events.
Seventh, publish post-assignment observations with restraint. Whether a prefix appears in routing can indicate activation but cannot prove beneficial use, customer geography or ownership. Use it as one signal, not as a revocation shortcut.
Eighth, report the cost borne by waiting members. Show applicable fee categories, IPv6-only discounts, waitlist administration cost and service received. Do not calculate a fictional cost per successful allocation by dividing the whole institution by a shrinking number of releases.
These records would improve both legitimacy and policy. Members could debate whether recovered space should remain a first-time allocation, whether the maximum should change, whether the line has become economically obsolete or whether priority should be protected for specific technical infrastructure. They would debate with denominators rather than anecdotes.
Scarcity turned visibility into a substantive member right
When supply was ample, a member could judge the registry mainly by the accuracy and speed of a completed request. Under terminal scarcity, the rejected and waiting cases became equally important. The institution's distributive power sat in order, information requests, phase interpretation, related-entity treatment and the timing of recovered releases.
LACNIC recognized part of this change by exposing queue position and approval timing. That was not cosmetic. It reduced the possibility that a request could be silently overtaken and gave applicants information for planning. The mistake would be to treat that achievement as complete transparency.
A visible line can still hide its gate. It can show admitted applicants while omitting those returned for information. It can show current positions while deleting attrition. It can show tickets while omitting requested quantity. It can show the central registry while excluding national channels. It can show supply while leaving the unmet-demand denominator unknown.
The proper conclusion is not that the queue was illegitimate. The phased policy preserved some entry after ordinary allocation had become impossible. It constrained block size, excluded prior recipients from the final reserve and later placed future recoveries behind quarantine. Those were real conservation and transition choices.
The conclusion is that conservation became allocation of economic opportunity. Once a /22 could avoid a market purchase, support customers or preserve independence from an upstream provider, queue administration carried consequences beyond clerical service. Membership accountability had to expand accordingly.
The last immediately available block did not make LACNIC irrelevant. It made the registry's information role more important and its rationing role more visible. The institution could no longer demonstrate fairness by proving that assigned addresses were unique. It also had to prove that conditional access was ordered consistently, that exclusions had reasons and that members could observe the demand the policy left unmet.
An honest exhaustion account therefore ends with two numbers, not one. The first is zero: the immediately available pool on 19 August 2020. The second is the number of qualified organizations still waiting, withdrawing or paying for alternatives. LACNIC documented the first with precision. Its next accountability task is to keep the second from disappearing.