Summary
- AFRINIC's Phase 2 queue makes final IPv4 allocations a test of timestamps, completeness, payment discipline and appealable process rather than a simple engineering request.
- Consider a small access operator in the AFRINIC service region.
The queue begins where engineering stops
Consider a small access operator in the AFRINIC service region. Its engineering case is not extravagant. It is not asking for a vanished era of generous IPv4 supply to return, and it is not pretending that a final pool can support decades of growth. It wants a block small enough to fit within AFRINIC's Phase 2 ceiling: perhaps a /22, perhaps less, enough to add broadband customers, attach enterprise sites, run public-facing services, or reduce the worst pressure on carrier-grade NAT. The technical story may be ordinary. The economic story begins when ordinary need meets a queue.
Under AFRINIC's published exhaustion procedures, Phase 2 requests are submitted in tickets and handled on a first-come, first-served basis. That phrase sounds simple until it is connected to the next rule: complete applications proceed to evaluation, while incomplete applications are handled case by case until the missing information is supplied and the file is considered complete. Final approval is not a continuously flowing act. AFRINIC describes a weekly final-approval rhythm, normally on Friday and on Thursday if Friday is a public holiday. The pre-approved list is sorted by the date on which each ticket became a complete application, from oldest to newest. The approved prefix is then reserved and an invoice generated. The reservation can last up to 45 days while payment is awaited; for new members, the signed Registration Service Agreement must arrive in the same window. If payment and the agreement do not arrive, the prefix returns to available inventory and the applicant must file again.
In that chain, the decisive question is not simply whether the network can use the addresses. It is whether the file became complete before another applicant's file, whether the evidence satisfied review quickly enough, whether payment cleared in time, whether corporate paperwork moved before the reservation expired, and whether the applicant could absorb the cost of delay. The scarce unit is not only the /24, /23, or /22. It is a place in time.
That is why AFRINIC's waiting list should be analysed as an economic institution. A queue is not the absence of allocation. It is a method of allocation that suppresses a direct price and substitutes time, paperwork, eligibility, proof, and administrative sequence. The queue can be legitimate. It can protect the final fragments of a regional pool from bulk capture. It can give smaller networks a way to obtain scarce addresses without bidding against better-capitalised buyers. It can make the registry look orderly during exhaustion. But it can also create delay cost, favour applicants with better paperwork capacity, encourage strategic filing, push operators into leasing and transfer markets, and turn calendar priority into a form of capital control.
AFRINIC's case is especially sharp because the region's remaining IPv4 pool is small, its public governance history has been stressful, and the market outside the registry pool prices IPv4 far above ordinary administrative fees. A 2021 Internet Governance Project analysis described AFRINIC as having held only about 2% of global IPv4 space and placed the registry's disputes in a wider setting of scarcity, market value, and institutional stress. Those claims should be read as public analysis, not as the last word on every contested event. They nevertheless identify the right economic setting. When scarce addresses have a market price, a waiting-list allocation at registry terms is a valuable claim. Whoever controls the queue controls access to that claim.
The narrow case for the registry is therefore strong but limited. AFRINIC must keep the allocation record coherent, avoid duplicate recognition, conserve a finite pool, test need, and process applications in a predictable order. It must not turn the queue into hidden discretion. A waiting list is tolerable only when it is transparent, auditable, appealable, and insulated from discretionary acceleration. Without those safeguards, first-come first-served becomes less like neutral sequencing and more like capital control by calendar.
A waiting list is a price system without prices
Economists often describe rationing as what happens when a scarce good is not allowed to clear through price. The good still goes somewhere. It is just distributed by another scarce currency: time in line, paperwork, eligibility, political connection, lottery, merit score, committee approval, purchase history, employment status, or physical presence. The currency chosen by the institution decides who pays and how the payment is hidden. A waiting list looks fair because everyone can see the line. It is less fair when the cost of entering, staying in, and reaching the front differs sharply across applicants.
IPv4 exhaustion creates exactly this problem. The addresses are scarce in a literal numerical sense. IPv6 deployment changes the long-term architecture of the internet, but it does not erase the medium-term need for IPv4 compatibility across customers, platforms, government systems, banking, access networks, hosting, security equipment, and legacy applications. When the registry's free pool is nearly gone and the transfer market prices addresses at meaningful levels, an allocation from a residual pool is not merely a technical input. It is a subsidised input distributed by rule.
AFRINIC's Phase 2 rules make that implicit subsidy small in block size but large in consequence. The minimum allocation or assignment is /24, and the maximum is /22 per allocation or assignment. A /22 is only 1,024 IPv4 addresses. For a national mobile operator, it is tiny. For a new local ISP, a specialised enterprise provider, a hosting operator, an exchange-related network, or a firm trying to avoid more NAT layers, it can still matter. The gap between registry terms and market alternatives gives the queue its value. If applicants could obtain the same addresses instantly at the same cost elsewhere, the line would be a clerical nuisance. Because they cannot, the line becomes an economic arena.
The queue also changes the price signal. In an open market, rising demand raises price, which encourages holders to sell, buyers to economise, and investors to fund substitutes. In a waiting list, demand raises waiting time, proof burden, and frustration. Instead of asking whether the address is worth the price, the applicant asks whether it can endure the process. That favours organisations with administrative capacity, cash buffers, legal support, and planning departments. It penalises firms whose growth arrives as a customer contract, a school connectivity award, a public-service deployment, or an enterprise project with a deadline. The price is paid in delay and uncertainty rather than in a posted invoice.
This does not mean a registry should auction the last pool or abandon need-based policy. The registry tradition is built around uniqueness, registration, aggregation, conservation, and fair access. A residual pool is not a commodity warehouse. Yet refusing to recognise the economic cost of a queue makes the policy weaker. It encourages officials and participants to say "no one paid more" while ignoring the firms that paid through lost customers, leased addresses, extra NAT systems, delayed service launches, or capital tied up in uncertain planning.
The right question is not whether a waiting list is a market. It is what hidden prices the waiting list imposes, who is best able to pay them, and whether those prices support the public reasons for the policy. If the cost of queue access falls mainly on weaker operators, the list can reproduce inequality while appearing neutral. If the queue is legible and bounded, it can ration scarcity without becoming arbitrary. That is the institutional test.
Phase 2 turns small blocks into institutional leverage
AFRINIC's public exhaustion page says the region is in Phase 2. In policy terms, Phase 2 begins when no more than a /11 of non-reserved IPv4 space is available in the final /8. During this phase, each allocation or assignment is bounded between /24 and /22. Members requesting additional IPv4 space must also show that at least 90% of all IP space delegated to them by AFRINIC is used efficiently and must satisfy the relevant contractual-obligation check. Those rules compress the old allocation model into a tightly rationed regime.
The compression matters. A maximum of /22 prevents a single successful request from swallowing too much of the residual pool. It also changes expectations. Applicants know that even a successful request will not solve medium-term growth. The queue is therefore not a route back to abundance. It is a temporary relief valve. That relief can be worth pursuing, but it cannot remove the need for transfers, leasing, CGNAT, IPv6 deployment, renumbering discipline, or tighter address planning. A queue allocation buys time; it does not buy independence from scarcity.
Small maximum allocations can be defended as fairness. They limit the prize. They reduce the incentive for one applicant to file an enormous claim. They make it easier to explain that the remaining pool serves many networks rather than a handful of large buyers. They also make the queue more politically visible. Each /22 resembles a seat in a small waiting room. When a seat is given to one applicant, it cannot be given to another. Administrative order therefore carries distributional consequences.
Phase 2 also concentrates leverage in procedural definitions. In an abundant pool, a delay in proof, payment, or review may be annoying but not decisive. In a thin pool, the moment a file becomes complete can decide whether the applicant receives addresses before inventory tightens further. The same rule that once managed paperwork becomes a rule for allocating scarcity. A completeness decision is no longer a clerical milestone. It is a timestamp with economic value.
The 90% efficient-use threshold adds another layer. It protects the residual pool from applicants that already hold underused AFRINIC-administered space. But it also makes historical record quality a condition of queue access. A firm with clean internal address management can prove its case quickly. A firm with older systems, inherited networks, acquired subsidiaries, customer privacy constraints, or fragmented records may have real use but slower proof. The threshold is therefore both a conservation rule and an administrative-capacity test.
AFRINIC's policy manual frames IPv4 allocation around actual need and immediate use, and warns against stockpiling and ordinary reservation of address space. Those principles fit scarcity. Yet they can become blunt if the review does not distinguish between waste and operational reserve. Networks need buffers for routing structure, enterprise commitments, disaster recovery, public services, customer migration, and security segmentation. A registry can require evidence, but it must not pretend that every quiet address is idle or that every planned address is hoarded.
The manual's special categories underline the point. Internet exchange and anycast provisions show that policy already recognises uses whose value is not measured by filling every address with a separate customer. The same manual also records that if a reserved /12 remains unused when the remaining available space has been allocated, it is to return to the AFRINIC pool under Phase 2 conditions. Scarcity policy therefore needs explicit categories, return conditions, and visible release rules.
Phase 2 is thus not merely a smaller address pool. It is a different institutional condition. The registry's routine acts have more leverage because the alternatives are costly. The smaller the maximum allocation, the more important it becomes that the line is clean, the rules are known, and the timestamps mean what they say.
Completeness becomes the scarce asset
The most important word in AFRINIC's queue design may be "complete." Public exhaustion material says complete applications proceed to evaluation, while incomplete applications are handled with applicants until the needed information has been supplied and the applications are considered complete. Later, the pre-approved tickets are sorted by the date they became complete. That design tries to prevent an applicant from gaining priority by filing a thin or speculative ticket. It also creates a new scarcity: completeness itself.
Completeness sounds neutral because every applicant can, in theory, supply the same categories of evidence. In practice, the ability to assemble a complete file is unevenly distributed. A mature operator may have IP address management software, updated customer assignment records, network diagrams, board approvals, payment processes, named corporate signatories, technical staff used to registry requests, and lawyers who can review the agreement quickly. A smaller provider may have accurate operational knowledge but less formal documentation. A public-sector or university network may have slow procurement and authorisation. A cross-border group may need documents from several subsidiaries. A startup may have demand in the form of signed customer letters but little historical address estate to show.
The rule that incomplete applications wait until complete is still defensible. Without it, queue order would reward placeholders. Applicants would file early and fill in details later, turning the line into a race of speculative claims. That would be worse. The hard part is ensuring that "complete" has a published, stable meaning and that applicants can know in advance what will qualify. If completeness changes with reviewer preference, the queue is no longer first-come first-served. It is first-to-satisfy-an-uncertain-reader.
Completeness also has a time value. Suppose two operators submit tickets in the same week. The first has real need but misses a corporate authorisation letter. The second has slightly later demand but a perfect file. If the second file becomes complete first, it can move ahead. That outcome may be fair if the first applicant failed a clear rule. It is less fair if the missing item was not published, if the applicant could not know the expected format, or if the reviewer accepted a similar item from another applicant. A queue's legitimacy depends on the discipline of these small decisions.
Because completeness is valuable, it encourages pre-investment. Operators that expect future scarcity will prepare evidence files before they need addresses. They will clean assignment data, update contacts, standardise internal records, arrange signatory authority, and make sure invoices can be paid quickly. That is good when it improves record quality. It is less good when it turns policy participation into a compliance arms race. The firms best able to carry idle administrative capacity are not always the firms with the strongest marginal need.
The registry can reduce this distortion by publishing a checklist, examples, deficiency categories, response windows, confidentiality options and anonymised statistics on why files are returned as incomplete. It should maintain a tamper-evident timestamp for ticket submission, first review, deficiency notice, applicant response, completeness determination, pre-approval, final approval, invoice issue, payment receipt, and prefix release or return. Such detail may sound excessive for a /22. It is not excessive when the /22 is scarce and the queue itself is the allocation mechanism.
Completeness is where administration turns into economics. The file that gets to "complete" first has converted paperwork into priority. That conversion can be legitimate only when the conversion rate is known.
Time order rewards preparation, not merely need
First-come first-served appears to avoid judgement. It does not ask the registry to decide whether a rural ISP deserves a block more than a datacentre, whether a school network outranks a hosting provider, or whether a government service should move ahead of an enterprise access project. It says: if you are eligible and complete, your place is determined by time. The appeal of the rule is obvious. It is simple, explainable, and less exposed to open political bargaining than discretionary prioritisation.
But time order is never only time order. It rewards readiness. The applicant that knows the rules, monitors the pool, prepares evidence, files early, answers deficiency questions quickly, and pays within the reservation window gains an advantage over the applicant that discovers scarcity only when a customer contract arrives. That advantage may reflect good management. It may also reflect unequal capacity. A queue is a competition in foresight.
In some markets, that is acceptable. The operator that plans ahead should not be punished for competence. AFRINIC's policy goals include conservation and fairness, not rescue from poor planning. If an applicant waits until the last moment, cannot show efficient use, and cannot complete paperwork, the queue should not be rearranged around it. The risk arises when preparation becomes detached from real deployment. If applicants file as soon as they can assemble a plausible case, not when the network truly needs addresses, the queue can become a reservation system by another name.
AFRINIC's rules try to control that risk through need evaluation and the 90% efficient-use requirement. The policy manual's insistence on actual need and immediate use is meant to prevent long-term stockpiling. Yet the economic incentive is clear: when the pool is small and the queue is valuable, waiting until need becomes urgent is risky. Rational operators will file earlier, document more aggressively, and keep future plans ready. The line between prudent preparation and strategic early claiming becomes hard to police without introducing discretion.
This is one of the paradoxes of non-price rationing. The institution suppresses bidding for the address block, but it creates bidding for administrative readiness. Firms bid in staff time, lawyers, consultants, compliance systems, documentation, internal planning, and management attention. The richer firm can often pay that bid more easily than the poorer one. The queue therefore reduces one inequality, the ability to pay a market price, while increasing another, the ability to pay process cost.
The solution is not to abandon time order. It is to make time order honest about its limits. AFRINIC should distinguish ticket submission time from completeness time and explain which one controls in each stage. It should publish whether deficiency responses preserve original position, under what conditions a file loses priority, and how much time an applicant has to cure a missing item before another complete file moves ahead. It should also disclose aggregate queue data without revealing sensitive applicant information: number of pending requests, oldest complete ticket, oldest incomplete ticket, average time to first response, average time to completeness, average time to final approval, and number of prefixes reserved but not yet paid.
Time order is a good servant and a poor mythology. It is useful because it limits overt favouritism. It is dangerous if it lets the institution pretend that all applicants face the same clock.
The invoice clock is part of the rationing system
The 45-day reservation window is not a minor accounting detail. It is the bridge between an approved request and an actual allocation. AFRINIC's public process says that once a ticket is approved, the resource is reserved and an invoice generated. The approved prefix is held for up to 45 days while payment is awaited. For a new membership application, the signed Registration Service Agreement must also be sent within the same period. If payment and the agreement do not arrive, the prefix returns to available inventory and the applicant must submit a new request.
That rule protects the pool from indefinite reservation. Without a deadline, an approved applicant could hold scarce addresses while deciding whether to proceed, finding financing, negotiating with customers, or waiting for a business condition to improve. The queue behind it would suffer. A 45-day clock forces commitment. It also returns unused reservations to the pool rather than letting them become dormant claims.
Yet payment timing can be a hidden allocation filter. Many African operators live with currency friction, bank delays, tax documentation, foreign-exchange controls, slow public procurement, and board-authorisation cycles. A fee that is modest for one applicant can be administratively difficult for another. If the invoice arrives near a public holiday, fiscal year-end, grant disbursement date, or procurement deadline, the operational ability to pay may decide whether the applicant keeps its place. The address block is rationed not only by need and completeness but by payment machinery.
This is especially relevant for new members. They must not only pay but also complete the signed agreement. Corporate authority checks can take time. A smaller firm may need a director to sign in another city. A public institution may require ministerial or procurement approval. A company operating across borders may need notarisation, translations, tax certificates, or board resolutions. The rule cannot wait forever, but the registry should recognise that the agreement clock and payment clock together produce a real barrier.
The invoice clock also creates a secondary queue dynamic. Prefixes reserved but unpaid are unavailable to other applicants. If many approvals sit inside the 45-day window, the visible pool may overstate what can be issued immediately. Applicants behind the line face uncertainty: a block ahead may be paid, returned, or delayed by correspondence. If returned, it may create new opportunity. If paid at the last moment, the line tightens. This makes queue transparency even more important. Aggregate reporting on reserved-but-unpaid inventory, expirations, and returns would help members understand whether the queue is moving or merely holding inventory in suspense.
Payment defaults can also create stigma. An applicant that fails to pay within the window must file again. That is sensible as inventory discipline, but it should not become an informal mark against future eligibility unless the rules say so. Non-payment may reflect changed business need, banking delay, or inability to complete documents, not bad faith. Conversely, repeated speculative approvals followed by non-payment can waste review capacity and hold prefixes. The registry needs defined treatment for repeated failures, not improvised irritation.
The 45-day rule is therefore an economic instrument. It converts approval into commitment, prevents dead reservations, and keeps inventory moving. It also shifts risk to applicants with weaker payment infrastructure. If the queue is to be legitimate, the invoice stage must be as auditable as the eligibility stage.
Delay is paid in leasing, NAT and deferred customers
The cost of waiting rarely appears in the registry's accounts. It appears in the operator's network. If a requested /22 is delayed, the operator must still serve customers, meet contracts, handle abuse reports, support applications, and keep public services reachable. Scarcity does not pause while a ticket is incomplete or a prefix is reserved. The applicant pays for delay through substitutions.
The first substitution is leasing. An operator can rent addresses from a holder or intermediary, often faster than waiting for a registry allocation or transfer. Leasing can be a rational response to temporary need, uncertain growth, or a queue whose timing is not predictable. It can also create dependency. The lessee may not control reverse DNS, routing security settings, geolocation records, abuse contact reputation, or termination risk as fully as it would with registry-recognised space. Contract terms must handle renewal, price escalation, sub-use, complaints, blacklisting, and migration. The queue's delay therefore becomes private contract risk.
The second substitution is transfer purchase. A firm can buy IPv4 rights in a recognised transfer market, subject to applicable registry rules and counterparty diligence. That may provide stronger control than leasing, but it requires capital, title confidence, settlement discipline, and time. If AFRINIC-administered transfers carry uncertainty, the buyer may look outside the region where policy allows, or it may discount AFRINIC-linked space. Again the queue is not separate from the market. It pushes demand into the market, and market conditions feed back into how valuable queue access becomes.
The third substitution is carrier-grade NAT. CGNAT is often technically workable and economically painful. It allows many users to share public IPv4 addresses, conserving scarce space. But it imposes costs in port management, logging, lawful-access response, application failure, customer support, enterprise exceptions, reputation spillover, and security troubleshooting. When a queue delays public addresses, the operator may add NAT capacity sooner or run it harder. Customers may see broken gaming, voice, VPN, hosting, payment, or remote-access use. The hidden price of the queue is then paid by users who never know why their network behaves differently.
The fourth substitution is delay of service. The operator may postpone a customer launch, decline a contract, slow a regional rollout, avoid a product that needs public addresses, or push customers toward cloud platforms with bundled public identity. The economic cost is lost opportunity. The public cost can be worse if the delayed service involves schools, clinics, local enterprise, public administration, emergency systems, or rural connectivity. A /22 is not enough to transform a national internet economy, but its absence can still block a local deployment.
The fifth substitution is architectural compromise. Engineers may fragment addressing plans, stretch old pools, assign addresses more densely than is healthy, accept uglier routing, or postpone cleanup because the next block is uncertain. These compromises accumulate. They make the network harder to operate, harder to audit, and harder to migrate.
The registry cannot eliminate all these costs. IPv4 exhaustion is real. But it should measure the queue with these costs in mind. Average review time, completeness delay, approval cadence, unpaid reservation rates, and appeal duration are not mere service metrics. They are economic signals. A slow queue imposes a tax through leasing, NAT, transfers, and lost customers. A fast but careless queue misallocates scarce space. The institutional challenge is to minimise delay that does not improve the decision.
Queue gaming is rational when the prize is underpriced
Every rationing system creates gaming. This is not a moral surprise; it is a predictable response to value. If a scarce input is available through a queue at below-market cost, applicants will study the rules. They will file earlier, improve paperwork, time requests, divide need, coordinate subsidiaries, seek repeated small allocations, maintain just enough utilisation to qualify, or choose leasing and transfers while preserving a queue position. Some behaviour is legitimate planning. Some is opportunistic. The difference is hard to define after the fact, which is why the queue must be designed against predictable incentives.
AFRINIC's Phase 2 maximum of /22 reduces the reward from any single request. That helps. It is harder for one ticket to capture a large share of the pool. But the policy manual also states that during exhaustion there is no explicit limit on the number of times an LIR or end user may request IPv4 address space, while each request remains subject to the relevant criteria. This creates a sequence problem. An applicant may not be able to get much at once, but it may plan a series of requests as utilisation and need are shown.
Serial requests are not inherently abusive. A growing operator may need exactly that path: use a small allocation efficiently, return with evidence, receive another small block, and continue. Such a pattern fits conservation better than one large speculative allocation. Yet serial requests can also reward firms that are good at cycling through proof, while applicants with sporadic but urgent demand wait behind them. It can encourage address planning aimed at qualifying for the next request rather than minimising total cost.
The completeness rule creates another gaming frontier. If priority depends on the date a file becomes complete, applicants will ask what the minimum complete file looks like. They may submit highly curated evidence that passes formal criteria while leaving broader context obscure. They may hold optional details until asked, preserving speed. They may use consultants who know reviewer preferences. Again, some of this is sensible. A clear, efficient application should not be punished. But if knowledge of unwritten expectations becomes valuable, insiders gain.
Payment windows can be gamed too. If an approved prefix is reserved for up to 45 days, an applicant can in effect hold a scarce claim while deciding whether to pay. The rule limits the hold, but the hold still has value. Repeated failures to pay can waste time for others. The registry needs a defined response, such as reporting aggregate unpaid-return rates and applying clear rules for repeated non-completion, without turning one failed payment into permanent suspicion.
Queue gaming can also occur through affiliated entities. A group may have several legal entities, each with its own need. Some are real operating companies; others may be shells. A registry must distinguish legitimate corporate structure from artificial fragmentation. That requires clear rules for related parties, common control, downstream assignment, merger history, and evidence of independent networks. Without clarity, the registry either lets fragmentation capture inventory or responds with broad discretion that chills legitimate groups.
The lesson is simple. When the prize is underpriced, gaming is not an anomaly. It is the rationing system's shadow market. A good waiting list does not assume virtue. It narrows the gaming surface, publishes the rules, logs decisions, and treats similar cases similarly.
Secondary markets become the safety valve
A waiting list does not abolish the IPv4 market. It often makes the market more important. When the queue is slow, uncertain, or capped at small blocks, operators look to leasing and transfers. When leasing and transfers are costly or uncertain, the queue becomes more valuable. The two systems are not opposites; they are connected vessels.
The Internet Governance Project's 2021 analysis placed the AFRINIC crisis in precisely this market setting. It noted that IPv4 addresses had developed a transfer-market price, moving from roughly $8 per address in 2017 to about $30 per address by 2021 in the cited market context, and that a /16 could be worth around $2 million. Those figures are not needed to compute a current price here. Their significance is that a registry allocation at administrative terms can be materially different from a market purchase. That difference creates arbitrage pressure around any residual pool.
Leasing is the quickest safety valve. It lets an operator obtain usable public addresses without waiting for permanent transfer or registry allocation. It also creates a private layer of governance. The lease contract must decide who controls routing authorisation, reverse DNS, abuse contact, reputation cleanup, geolocation correction, termination notice, customer continuity, and emergency cooperation. If the registry queue is trusted and timely, leasing can be a bridge. If the queue is opaque, leasing becomes a structural substitute.
Transfers are the more capital-intensive safety valve. They can move addresses from lower-value to higher-value uses and reveal market price. But transfers depend on registry recognition, title confidence, clean records, settlement mechanisms, and predictable review. If the registry is perceived as able to delay, question, or reverse transfers unpredictably, the transfer market discounts the space. Operators then keep more addresses than they need, buyers hesitate, and brokers price in institutional risk. A queue that was intended to conserve the pool can indirectly reduce liquidity elsewhere.
This is where market-oriented critics of need-based control have a point, even if one does not accept every conclusion. Their warning is that scarcity without price does not make value disappear. It relocates value into permission, delay, and administrative leverage. When an address cannot move easily through market mechanisms, the institution controlling recognition becomes the place where value is contested. The result can resemble a ration coupon: not ownership in the full economic sense, but conditional access to a scarce input.
The registry cannot simply become a price-taking exchange. It has public coordination responsibilities: uniqueness, registration accuracy, policy consistency, and continuity. But it should understand that a waiting list with weak transparency does not protect the market from speculation. It may increase speculation by making official allocation feel arbitrary and by pushing unmet demand into opaque leasing arrangements. Conversely, a predictable queue can make the market healthier by clarifying which needs may be met from residual inventory and which needs must be priced elsewhere.
The safety-valve role also means the queue should not be judged only by how many addresses it issues. It should be judged by how much avoidable uncertainty it adds to the wider address economy. A queue that issues small blocks predictably may coexist with transfers and leasing. A queue that issues small blocks unpredictably becomes one more risk premium.
Transparency is the queue's currency
A waiting list runs on trust. Price rationing has a harsh form of transparency: the price is posted, even if many cannot pay it. Queue rationing must create its own transparency. Applicants need to know how many requests are ahead, which timestamps matter, what evidence is required, how long review usually takes, how many prefixes are reserved but unpaid, how often reservations lapse, and how disputes or appeals affect the line. Without such information, the queue becomes a rumour market.
AFRINIC's public materials provide some important mechanics. They describe ticket submission, first-come first-served handling, completeness review, weekly final approval, date-of-completeness sorting, 45-day reservation, the Phase 2 /24-to-/22 range, the 90% efficient-use requirement, and the return of unpaid prefixes to inventory. These are valuable disclosures. They should be the beginning of queue transparency, not the end.
The next layer is operational reporting. A registry does not need to publish applicant names or sensitive details to make the queue auditable. It can publish aggregate statistics: total active requests by status; number of incomplete tickets awaiting applicant response; number of complete tickets awaiting evaluation; oldest complete-ticket date; median and 90th-percentile time from submission to first response; median and 90th-percentile time from submission to completeness; median time from completeness to approval; count of prefixes under 45-day reservation; count returned after non-payment; and total addresses issued under Phase 2 over a reporting period.
Transparency also protects staff. Hostmasters reviewing scarce requests are in a difficult position. They must test need, reject weak files, ask for sensitive evidence, and handle applicants with real business pressure. A visible process reduces pressure for informal updates, personal escalation, and accusations of favouritism. It allows staff to say that the line is governed by recorded status, not by who complains most loudly.
The audit trail should be tamper-evident. Each status change should have a timestamp, a reason code, and the responsible role. Applicants should be able to see their own file history. The community should see aggregate movement. Appeals bodies or independent auditors should be able to examine specific disputes. This is not technological glamour. It is basic economic infrastructure for a non-price rationing system.
Transparency must include negative facts. If the registry has no prefixes left in a category, say so. If the reserved /12 is not yet available because the return condition has not occurred, explain the condition. If payment defaults are holding inventory, show the aggregate amount. If a court order or governance interruption affects processing, state the operational effect without theatrical language. Uncertainty is less damaging when it is named.
The queue's currency is not goodwill. It is verifiable process.
Appeals prevent time order from becoming discretion
Every queue needs an appeal path because every queue makes mistakes. A file may be marked incomplete based on a misunderstanding. A payment may be delayed by a banking error. A reviewer may apply an evidence expectation inconsistently. An applicant may be treated as related to another entity when the relationship is operationally separate. A use category may be misread. A timestamp may be wrong. In a scarce pool, such errors have economic value. Correcting them after the addresses are gone may be impossible.
An appeal path does not mean every disappointed applicant gets another chance to argue need from scratch. It means the applicant can challenge a defined decision: incomplete status, eligibility rejection, loss of queue position, expiry of reservation, refusal of evidence format, related-party treatment, or other process defect. The appeal should ask whether the published rule was applied correctly and whether similar cases were treated similarly. It should not become a discretionary shortcut around the line.
Appeals are often criticised as slowing the queue. They can, if poorly designed. But the absence of appeal can slow the system more. Applicants that believe the process is arbitrary will escalate informally, litigate, lobby, or flood staff with correspondence. Public trust erodes. Every future rejection becomes more contentious. A bounded appeal is a pressure valve that keeps disputes inside procedure.
The design should protect both the appellant and the queue behind it. If an applicant appeals loss of a specific prefix after the 45-day window, should that prefix remain reserved? If yes, how long? If no, what remedy exists if the appeal succeeds? If an applicant challenges an incomplete determination, does its original completeness date change if the appeal succeeds? If a rejection is overturned, does the applicant regain its former place or enter the current line? These questions must be answered before the dispute occurs. Otherwise appeal itself becomes a new arena for discretion.
There is also a need for separation of roles. The person who made the contested decision should not be the final reviewer of the appeal. Staff expertise is essential, but scarcity decisions need procedural independence. The appeal body or reviewer should have access to the full file, status history, deficiency notices, applicant responses, and comparable anonymised cases. Its decisions should be reasoned and, where possible, summarised publicly in redacted form so future applicants understand the rule.
AFRINIC's broader governance history makes this more than good administration. Public reporting has described court actions, bank-account freezes, board and election stress, receivership, and intense member conflict. A registry under such conditions should avoid putting more weight on informal trust than necessary. Appeals turn trust into procedure. They also protect the registry by showing that adverse decisions are reviewable before they become external disputes.
The point is not to judicialise every /24. It is to recognise that queue position is an economic interest. Where a decision can move that interest, there must be a way to test the decision without buying influence or starting a public war.
Discretionary acceleration is the queue's fatal weakness
The gravest risk in a waiting list is not slowness. It is selective speed. A slow queue that applies clear rules frustrates everyone. A queue that accelerates some applicants without auditable reason teaches everyone else that procedure is theatre. Once that belief spreads, every status update becomes suspicious, every delay becomes a possible favour to someone else, and every allocation is read as a political act.
Discretionary acceleration can take many forms. A file may receive faster first review. A deficiency may be phrased more helpfully for one applicant than another. A late document may be accepted without loss of position. A payment problem may receive informal patience. A completeness date may be interpreted generously. A related-party issue may be overlooked. A special use may be treated as obvious in one case and doubtful in another. None of these acts has to look dramatic. In a scarce pool, small procedural differences can decide outcomes.
This is why a first-come first-served system needs strict separation between legitimate expedition and improper acceleration. Legitimate expedition means increasing capacity for all files, publishing clearer checklists, automating status notices, holding regular approval meetings, and clearing old tickets according to rule. Improper acceleration means moving a particular applicant ahead because of status, pressure, relationship, political appeal, public narrative, or staff sympathy. The former improves the queue. The latter destroys it.
Emergency categories are especially dangerous if left vague. A registry may face requests tied to public safety, national infrastructure, health systems, exchange stability, or security incidents. Some may deserve special handling if policy provides for it. But if the ordinary queue is first-come first-served, emergency exceptions must be narrow, published, time-limited, and reported. Otherwise every applicant has incentive to frame its need as exceptional. Scarcity produces persuasive stories.
The same applies to governance pressure. A board, receiver, executive, or government-facing official may want to show that the registry is supporting a politically visible project. The temptation can be high in a region where connectivity, development, sovereignty, and institutional legitimacy are linked. But a registry's credibility depends on resisting case-by-case acceleration unless the policy permits it. The queue is not a development bank. It is a rationing mechanism for a finite address pool.
Discretion can also appear in the opposite direction: selective slowing. A controversial applicant may receive more questions, longer review, stricter evidence demands, or less tolerance for payment difficulty. If the registry has broader concerns about an applicant, it should use defined eligibility and compliance tools, not make the waiting list do hidden enforcement work. Slow-walking is as corrosive as jumping the line.
The remedy is radical mundanity. Publish the status categories. Record every change. Use reason codes. Maintain peer review. Separate completeness from final approval. Report aggregate timing. Allow appeals. Audit comparable cases. Keep political actors away from individual tickets. A waiting list should be boring enough that nobody can plausibly treat it as a lever.
Scarcity will always create pressure for exceptions. The institution's job is to make exceptions rare, visible, and rule-bound.
Governance stress raises the premium on boring process
AFRINIC's waiting-list economics cannot be separated from institutional context, but the context should be used carefully. The point is not to replay courtroom drama or treat every public allegation as settled fact. The point is that a registry with a stressed governance history needs stronger process because members will not supply trust automatically.
Public reporting has described AFRINIC as having faced allegations of historical misappropriation of IPv4 resources, intense conflict around Cloud Innovation, court-ordered restraints, a provisional freeze on bank accounts in Mauritius, and wider debate over the organisation's continuity and governance. The Internet Governance Project's 2021 account argued that AFRINIC held a small share of global IPv4 space and that disputes over remaining addresses were, in part, disputes over scarcity and market value. Whether one agrees with every judgement in that analysis, the operational lesson is plain: when scarce resources and institutional conflict meet, process design becomes economic policy.
In a calm institution, members may accept some opacity because they believe the staff culture is stable and the governance system will correct mistakes. In a stressed institution, the same opacity is interpreted less generously. Applicants ask whether delays reflect capacity problems, internal politics, litigation risk, undocumented priorities, or fear of future challenge. Staff may become more cautious, asking for more documents and moving more slowly. Applicants may become more aggressive, escalating earlier and documenting every exchange for possible dispute. The queue becomes heavier even if the formal rules do not change.
This creates a vicious circle. Governance stress reduces trust. Reduced trust increases the demand for procedural detail. More procedural detail can slow processing if it is bolted on manually. Slower processing increases frustration. Frustration feeds governance conflict. The way out is not charismatic leadership or slogans about community. It is systems that reduce reliance on personal trust: published criteria, recorded timestamps, independent review, aggregate reporting, and clear boundaries between allocation review and wider enforcement.
The queue should also be insulated from institutional survival incentives. A registry under financial, legal, or political pressure may be tempted to use scarce-resource administration to demonstrate relevance. It may emphasise control, review, and discipline to reassure some members or critics. But the waiting list's legitimacy depends on the opposite posture: narrowness. The registry is not distributing favours. It is applying a finite policy under exhaustion. Its authority grows safer when it does less by discretion and more by record.
Governance stress also affects applicants' alternatives. If members doubt registry continuity, they will value certainty elsewhere: leases with strong contracts, transfers in other regions where allowed, cloud platform addresses, or conservative network designs that reduce dependence on future AFRINIC allocations. That may be rational for each firm, but it weakens the registry's role as trusted coordinator. The queue then loses not only speed but meaning.
Boring process is not a low ambition. In a scarce-resource institution, boring process is legitimacy.
The queue can become capital control by calendar
Capital control usually means a rule that limits how money or assets can move. In the IPv4 economy, a waiting list can perform a similar function without using that language. It decides when a network may obtain scarce address capital from the residual pool. It decides whether the applicant's proof is acceptable. It decides whether delay forces the applicant into leasing or purchase. It decides whether an underpriced claim reaches the applicant or returns to inventory. The control is exercised through calendar and file status rather than through a posted prohibition.
This is why market-oriented commentary on IPv4 scarcity often attacks need-based control. The critique is that when price signals are suppressed, permission becomes the scarce resource. A firm may have customers, capital, engineering skill, and demand, but still need institutional approval to obtain or move addresses. In that setting, value can freeze. The address economy shifts from ownership and exchange toward application and waiting.
That critique should not be treated as a reason to abolish every registry rule. A unique-address system cannot function without a recognised public registry. Transfers must be recorded. Duplicate claims must be prevented. Fraud must be controlled. Residual pools must be distributed according to policy. But the critique is useful because it names the danger: a registry can slide from maintaining the ledger into controlling capital allocation. The waiting list is one of the places where that slide can happen quietly.
The clearest sign of capital control by calendar is when applicants cannot predict the cost of delay. If the queue has stable metrics, known evidence rules, visible status, and real appeal, delay is an operating risk. It can be planned around. If the queue is opaque, delay becomes unpriceable. Operators must over-hedge: lease more, buy earlier, hold more old space, avoid customers, or maintain larger NAT and support buffers. The economy pays for uncertainty.
Another sign is when completeness becomes a discretionary barrier. If a file is never complete because new questions keep arriving, the applicant is not waiting in a line. It is negotiating permission. If payment timing is treated harshly for some and flexibly for others, the invoice clock becomes a gate. If public-interest stories move some requests faster, the queue becomes a political allocator. If controversial applicants are slowed through extra proof demands, the waiting list becomes hidden enforcement.
The phrase "capital control by calendar" is deliberately severe. It does not mean every delay is abusive. It means that in a scarce IPv4 environment, time is a capital variable. A month of delay can require leased addresses, lost revenue, extra NAT capacity, or transfer-market exposure. The institution that controls the timing controls an economic choice.
AFRINIC can avoid this charge only by narrowing its queue power. It should be able to say: here is the rule, here is the timestamp, here is why the file is incomplete, here is how to cure it, here is the appeal path, here is the approval cadence, here is the reservation clock, and here is the aggregate queue. That is the difference between rationing scarcity and commanding capital.
What a legitimate waiting-list regime should publish
A credible waiting-list regime can be described in operational terms. It should publish the active policy basis for Phase 2 requests: minimum and maximum sizes, efficient-use requirements, good-standing checks, evidence categories, special-use categories, and any limits on repeated requests or related entities. These rules should be readable by a network operator before filing, not reconstructed from correspondence after a deficiency notice.
It should publish queue states with timestamps. A ticket might be submitted, awaiting first review, awaiting applicant response, under evaluation, complete, pre-approved, pending final approval, reserved pending payment, issued, returned after non-payment, rejected, appealed, or withdrawn. The exact labels matter less than consistency. Each state should have a meaning, a responsible role, and a possible next step. Applicants should not have to guess whether they are waiting for staff, waiting on themselves, or waiting for the weekly approval cycle.
It should publish aggregate metrics: number of requests in each state, oldest complete request, time to first response, incomplete-file rates, deficiency-cure time, complete-file waiting time, prefixes inside the 45-day reservation window, and returns after non-payment. It should also publish inventory categories: available inventory, reserved Phase 2 inventory, special reserves, prefixes reserved pending payment, prefixes under dispute, returned unpaid prefixes, and any reserve scheduled to return under policy.
Finally, it should publish completeness guidance, appeal rules and acceleration safeguards. Common missing items, acceptable evidence, confidentiality options, redaction standards and efficient-use calculations should be known in advance. Appeals should cover decisions that affect queue position. No ticket should move forward outside the published process without a recorded rule basis. This publication burden is the price of non-price rationing: the queue must be legible enough that the hidden price of time does not become hidden power.
The post-exhaustion bargain
The allocation-era bargain was simple enough to feel technical. A network justified need. The registry checked policy. Addresses were delegated. The public record was updated. The resource was scarce in theory, but the pool was large enough that many decisions looked administrative. Exhaustion changed the bargain. Now each residual allocation is a distributional act. The registry can still speak in technical terms, but the economic content is visible to every applicant that faces leasing quotes, transfer prices, NAT costs, customer deadlines, and bank paperwork.
AFRINIC's waiting list is therefore a test of post-exhaustion legitimacy. The institution does not need to prove that it can create IPv4 abundance. It cannot. It needs to prove that it can administer scarcity without converting registry authority into discretionary economic power. That means narrowing its role to the record, the policy, the queue, and the evidence needed for the immediate decision.
The bargain for members should be equally clear. Applicants cannot treat the final pool as a substitute for planning. They must keep accurate records, use previous delegations efficiently, prepare evidence, pay invoices quickly, sign required agreements, and accept that a /22 maximum is not a growth entitlement. They should not expect weak files to hold priority or unpaid reservations to last indefinitely. Queue legitimacy requires applicant discipline too.
The bargain for the wider market is that the registry will not pretend secondary markets are irrelevant. Transfers, leasing, and CGNAT are not failures of morality. They are predictable adaptations to scarcity. The registry's job is not to bless every private arrangement, but neither should it use the residual pool as leverage to deny market reality. If official allocation is small and slow, markets will absorb demand. The question is whether registry process adds certainty or uncertainty to that adaptation.
The bargain for governance is restraint. Scarcity will always tempt institutions to expand their mandate. A queue invites the institution to become a judge of urgency, virtue, regional worth, commercial purity, and political attractiveness. That temptation should be resisted. If the community wants priority categories, it should adopt them openly. If it wants first-come first-served, the institution should administer time order faithfully. Hidden priority is worse than declared policy.
AFRINIC's remaining IPv4 pool is too small to justify large institutional myths and too valuable to administer casually. The queue is the point where those facts meet. It can be a narrow mechanism for rationing the final pool with dignity: complete files, known timestamps, weekly approvals, bounded reservations, efficient-use evidence, and appealable decisions. Or it can become a calendar-shaped gate through which economic opportunity passes slowly, unevenly, and with too much room for interpretation.
The difference will not be decided by speeches about scarcity. It will be decided by the dull machinery of queue design. In final IPv4 allocation, dull machinery is exactly what legitimacy requires.