The registry layer becomes a balance sheet issue
AFRINIC is often described in technical language: a Regional Internet Registry, a nonprofit member-based organisation in Mauritius, a steward of internet number resources for Africa and parts of the Indian Ocean. Its own public materials say it distributes and manages IPv4, IPv6 and autonomous system numbers; its policy manual says number-resource policy is made through a bottom-up process; its exhaustion page explains how scarce IPv4 is rationed through "soft landing" phases. Those descriptions are accurate as far as they go. They are also incomplete. The institutional question raised by AFRINIC is no longer only whether an address registry keeps clean records. It is whether a registry with discretion over transfers, membership standing, regional use and ledger updates is now exercising capital control over productive network assets.
The phrase is deliberately economic. Capital controls do not always seize property. Often they restrict movement, impose permissions, regulate convertibility, add reporting duties, limit exit, or make a transaction valid only after administrative approval. The nominal owner or holder may still possess the asset. The practical value changes because the asset can no longer move freely to its highest-valued use. A bank balance trapped behind exchange controls is still a balance. A factory whose profits cannot be remitted is still a factory. A bond that can be sold only to approved domestic buyers is still a bond. Yet each trades at a discount because exit has been narrowed.
IPv4 addresses are not factories or bonds, and a registry is not a central bank. But the economic mechanism is similar enough to matter. A scarce IPv4 block sits inside live networks, customer contracts, allowlists, abuse histories, geolocation databases, firewall rules, reverse DNS, route objects, RPKI arrangements and service-level expectations. It can be leased, sold, financed indirectly, valued in transactions, pledged as part of a business plan, or used to support revenue many times larger than the registry fee attached to it. Once that is true, rules about who may transfer it, whether it may leave a region, which member may receive it, how use must be justified and when records may be changed are not merely clerical. They change the economics of the asset.
Lu Heng's public note "The Policy Mirror", published on heng.lu on 30 June 2026, gives the sharpest framing for this shift. It reads AFRINIC's policy architecture as a movement from ledger to control: a registry justified by uniqueness and contactability begins to speak in the language of stewardship, regional eligibility, proper use, needs assessment, revocation exposure and approved channels. The note is polemical in places, but its institutional-economics point is sober: once IPv4 becomes scarce capital, old allocation-era vocabulary stops being harmless. The same phrases that once rationed an unallocated pool can become restrictions on assets already embedded in productive systems.
Official AFRINIC sources provide the factual exhibits. The Consolidated Policy Manual defines the goals of the IPv4 registry system as uniqueness, registration, aggregation and conservation. It says public unicast IPv4 addresses must be globally unique, that assignments and allocations must be registered in the AFRINIC database, and that conservation means distribution according to actual need and immediate use. It also says assignments are for specific purposes and are not to be sub-assigned to other parties. Provider-aggregatable space is described as non-portable: if the downstream network changes provider, the address space should be returned and the network renumbered. Provider-independent space is treated as portable but expensive to route and unavailable for sub-allocation by the end user or LIR.
Those clauses are not eccentric if read as documents from the allocation era. Before exhaustion, a registry distributing a finite free pool needed criteria. Need, immediate use and anti-stockpiling rules could be defended as rationing devices. The trouble begins when the same ideas migrate from free-pool allocation into transfer-era asset governance. AFRINIC's own IPv4 exhaustion page says the region entered Soft Landing Phase 2 on 13 January 2020, after earlier notices in 2017 and 2019. In Phase 2, the maximum allocation or assignment is only a /22, and additional requests require efficiency thresholds and contractual checks. Africa's growth therefore cannot be supplied mainly by new AFRINIC allocations. It must rely on reuse, transfers, leasing, better utilisation, IPv6 where feasible, and commercial arrangements that bring idle addresses into service.
At that point, the registry's most important economic role is not distribution from abundance. It is governance of movement under scarcity. The difference is decisive. Distribution from a free pool asks who should receive something not yet held by an operator. Movement under scarcity asks whether an existing holder may redeploy, sell, lease, finance or transfer an already productive input. The first is administrative rationing. The second touches capital formation.
The instruments of control are mundane
Capital control at the registry layer does not arrive as a dramatic proclamation. It is built from ordinary clauses. A recipient must justify need before AFRINIC. A transfer source must be a rights holder recognised by AFRINIC and not involved in a dispute. A recipient must become an AFRINIC member, sign the Registration Services Agreement and accept current policies. Transferred legacy resources no longer remain legacy resources under the intra-region transfer section captured in AFRINIC's manual. Transfers require good standing, according to the fee schedule. The billing page describes a closure process for unpaid fees and ties membership fees to the aggregate resources held. Each rule can be defended in isolation. Together they create a permission architecture around scarce assets: eligibility, status, payment, geography and record change all meet at the registry desk.
The simplest control is the needs test. AFRINIC's manual says a transfer recipient must justify and demonstrate its initial or additional IPv4 usage according to policies in force. In a free-pool queue this may be reasonable. In a purchase or lease market it is a different claim. The buyer's need is partly revealed by willingness to pay and willingness to bear operational risk. A hosting company buying addresses for a product line, a cloud provider leasing blocks for customer workloads, a security vendor requiring clean dedicated space, an enterprise seeking portable continuity, or a broker structuring a bulk transaction will often know more about its future demand than a registry reviewer can infer from a ticket. When the reviewer has power to approve or deny the transfer, the need test becomes a gate on investment.
The second control is regionality. AFRINIC's current older transfer section is explicitly "within the AFRINIC region"; Lu Heng's public note says the 2026 transfer policy goes further by classifying resources by origin and status, treating some resources as regional, reserved, legacy or global, and making only certain categories movable out of the AFRINIC region. The official snapshot available locally confirms the older core: transfers are framed as a way for African organisations to obtain IPv4 after AFRINIC can no longer satisfy their need. That premise may sound developmental, but it embeds a regional border around a globally routable identifier. A route does not have a passport. A prefix used by a customer in Lagos can serve users in London; a block held by a Seychelles company can support customers in Asia; a cloud service can locate compute, customers, routing policy and corporate domicile in different places. The registry region is useful for service delivery. It is not the same thing as the geography of economic use.
The third control is member status. AFRINIC's public materials describe a member-based structure, fees, renewal invoices, good-standing requirements and a closure process. Membership is not merely a club subscription when the member's continued standing affects registry services, transfer processing, reverse DNS, RPKI, contact records or recognition of number resources. If a network treats its IPv4 block as operational capital, any uncertainty about membership standing becomes an uncertainty about the asset's usable life. In ordinary commerce that risk would be priced, insured, negotiated or diversified. In a registry monopoly it is harder to escape.
The fourth control is ledger discretion. The registry record does not carry packets, but it shapes the world around packets. WHOIS and RDAP records identify holders and contacts. Reverse DNS delegation depends on registry coordination. IRR and RPKI information influence route filtering and security decisions. Transfer records influence who lenders, buyers, lessees, counterparties and courts regard as the legitimate control party. When a registry refuses to update a record, delays a record, marks a resource disputed, withholds a related service or threatens non-recognition, it can change the commercial value of a resource even if the route still announces.
The fifth control is status conversion. The manual's statement that transferred IPv4 legacy resources will no longer be regarded as legacy resources is an example. Status conversion is a quiet but powerful mechanism. It tells holders that entering the AFRINIC record system, or accepting a transfer under its rules, may bring the resource under a policy environment that differs from the one in which the holder previously relied on it. That discourages inbound capital. A seller or buyer considering an AFRINIC-side transaction will ask whether the block can later leave, whether its status survives, whether future policies can attach new conditions, and whether the registry may interpret acceptance of service as surrender of prior protections.
Each instrument has a benign administrative face. Needs assessment sounds like fairness. Regional retention sounds like development. Good standing sounds like ordinary billing discipline. Ledger discretion sounds like data quality. Status conversion sounds like policy consistency. The economics, however, depends on combined effect rather than label. If these controls reduce exit, make approval discretionary, attach new risk to imported resources and allow policy to follow assets after allocation, they operate like capital controls over a registry-mediated asset class.
Why a regional transfer wall lowers value
The strongest argument for regional limits is also the most intuitive: Africa received a small share of the global IPv4 pool, African networks still need addresses, and unrestricted transfers might drain scarce resources toward richer foreign buyers. The Internet Governance Project's 2021 analysis of the AFRINIC crisis captured the background. AFRINIC came late to the RIR system, held only a small fraction of global IPv4, and for a time remained the last region with a relatively large pool available at nominal administrative fees while other regions had exhausted their pools. That gap between low registry fees and rising global market value produced arbitrage. It also produced political pressure to keep "African" addresses for Africa.
The problem is that a regional wall does not create more addresses. It changes who can bid for them and under what conditions. If an AFRINIC-registered block can be transferred only within the region, the holder faces a narrower market. A narrower market usually means a lower exit price, lower collateral value, lower willingness to discover and release idle supply, and a larger spread between official and unofficial arrangements. The block may still be routed globally; the value that cannot move through the official registry channel will try to move through leases, nominees, customer delegations, long-term service contracts or informal control structures. The registry then sees less of the real operational picture, not more.
Development policy often commits this error. It treats an asset's physical or administrative retention as equivalent to local benefit. But a region benefits from assets not only by keeping them inside a border; it benefits when assets can be valued, financed, pledged, leased and redeployed efficiently. If a small African ISP holds more IPv4 than it can use, a liquid market gives it capital. It can sell part of the block, lease part of the block, finance network upgrades, buy transport, improve service, or invest in IPv6 transition where that makes commercial sense. If the same block is trapped in a restricted market, the ISP's balance sheet is weaker. The asset becomes less useful because exit value is suppressed.
Regional retention can also make inbound supply less likely. The Internet Governance Project ended its 2021 piece with a blunt point: Africa's future Internet growth cannot be sustained by the tiny leftover portion of AFRINIC's IPv4 pool. Growth requires importing large numbers of IPv4 addresses from the market, relying more on IPv6, or both. A policy environment that tells outside holders they may lose status, face exit restrictions or become subject to open-ended local conditions discourages exactly that import. Capital that cannot leave will demand a discount before it enters. Some capital will not enter at all.
This is the poverty penalty inside registry protectionism. A rule intended to preserve resources for poorer or later-developing networks can reduce the value of the resources those networks already hold and raise the risk premium for resources they need to import. In a mature asset market, the poorer region would want more liquidity, not less: cheaper transaction costs, cleaner title-like reliance, objective transfer recording, transparent dispute handling, financing channels and leasing records that make idle supply visible. A regional embargo does the opposite. It makes scarcity more political and less productive.
The collateral analogy is important. IPv4 is not a mortgageable land title in the conventional sense, and registries resist property language. Yet businesses rely on IPv4 in collateral-like ways. A data-centre operator's revenue projections assume addresses can continue supporting customers. A cloud platform's sale value may include address inventory and associated customer continuity. A lender or investor reviewing an infrastructure company will ask whether a key input can be sold, transferred, leased or preserved through distress. A buyer of an ISP will care whether the addresses can move with the business. If policy makes those answers uncertain, the asset is discounted even when lawyers avoid the word "property".
In this sense the market is not waiting for theoretical clarity. It already prices registry risk. Lu Heng's public notes repeatedly argue that IPv4 has become scarce, priced and operationally embedded capital, while public LARUS materials market leasing partly as a way to place registry-layer risk upstream rather than inside the customer's operating company. One need not accept every claim in those commercial materials to see the signal. The existence of products built around continuity, leasing, first-party pools and registry-risk allocation shows that operators are already treating the registry layer as a material risk surface. A policy that increases that risk surface will not stop the market. It will alter the market's structure, usually away from transparent transfers and toward risk-arbitrage.
Leasing is not a loophole; it is a scarcity response
The leasing question exposes the weakness of allocation-era thinking. AFRINIC's manual says assignments must be made for specific purposes and not sub-assigned to other parties. That logic fits a world in which an LIR receives addresses for downstream distribution under a policy hierarchy and the registry wants to know why the resource is needed. In today's IPv4 economy, however, commercial delegation is ordinary. Hosts lease addresses to customers. Cloud providers assign addresses dynamically. Enterprises use providers' addresses in one context and bring their own addresses in another. Security vendors need temporary dedicated blocks. Mobile networks, SaaS platforms and content providers use address space across layered contractual relationships.
Calling all of that suspicious does not make it disappear. It makes the database less accurate. If a holder fears that admitting a lease or sub-delegation will trigger review, needs re-justification, regional-use questions or transfer constraints, the holder has incentives to keep the arrangement at the contract layer while leaving the public record thin. The registry may still see the original holder, but not the customer, lessee, routing intermediary, abuse contact or operational controller. A policy designed to preserve order then produces opacity, which is the opposite of a registry's job.
Public LARUS material makes the market response explicit. Its leasing page presents first-party IPv4 leasing as a way to obtain production capacity while keeping registry-layer contract exposure, policy risk, audit pathways and termination mechanics upstream. It distinguishes between buying addresses directly and leasing from a specialist provider that claims to absorb registry risk. Again, this is not neutral academic evidence; it is a commercial position from an interested actor. But interested actors reveal market pressures. If customers were indifferent to registry-side risk, there would be little reason to sell continuity as the product. The pitch exists because buyers and lessees increasingly understand that the invoice price of addresses is not the only cost. The real loss from failure can be renumbering, routing changes, firewall and allowlist rework, customer churn, abuse-history disputes, downtime and contract exposure.
A registry concerned with accuracy should therefore prefer recognised leasing to hidden leasing. It can require the holder of record to remain identifiable. It can allow optional publication of delegated contacts. It can preserve abuse-report contactability. It can record time-limited operational control where parties choose to disclose it. It can maintain RPKI and reverse-DNS transition mechanisms. It can mark disputes without pretending to adjudicate every contract. Those are registry functions. They improve truth.
Prohibiting or stigmatising leasing tries to solve an economic problem with vocabulary. Scarcity creates a price for use. A party with idle addresses can earn recurring revenue by letting another party use them. A party that needs addresses but does not want full registry exposure can rent capacity. Leasing allows smaller networks to obtain capacity without buying an entire block, and it allows holders to monetise underused inventory without permanent sale. None of this is inherently a uniqueness problem. A leased address can be unique, accurately recorded, reachable through a proper abuse contact and supported by valid route authorisations. The risk comes from opacity, fraud, duplicate control, bad contact data or broken security assertions - not from the fact that the commercial arrangement is a lease.
The African development argument can be turned around here too. If local operators lack capital to buy large blocks at global market prices, leasing is one of the mechanisms that can bring usable supply to them. A registry that treats leasing as morally suspect may push African networks toward worse options: small final-pool allocations, NAT complexity, grey-market delegations, or dependence on providers whose address policy becomes customer lock-in. A registry that recognises leasing can make the market safer. It can ask for accurate records and objective safeguards without pretending the market should not exist.
The Cloud Innovation dispute showed the downside of discretionary review
The AFRINIC-Cloud Innovation conflict is not the whole story, but it is the clearest stress test. Independent reporting by the Internet Governance Project in August 2021 described the dispute as arising after AFRINIC's attempt to clean up after past problems, including the reported theft and sale of IPv4 space by a former senior staff member. According to that account, AFRINIC audited address registrations and targeted Cloud Innovation because of its business of leasing addresses, including to out-of-region customers. AFRINIC raised concerns about discrepancies between registered usage and actual countries of use, consistency between justified need and current use, and service origination within the AFRINIC region.
The same IGP account records why the dispute became economically explosive. AFRINIC asked for detailed information about Cloud Innovation's use of addresses and said it could determine whether to terminate the Registration Service Agreement and reclaim IPv4 resources. Cloud Innovation saw that as an existential threat because the addresses supported thousands of customers. AFRINIC's letter, as described by IGP, also disclaimed liability for loss or damage arising from the notice or any action taken under it. Cloud Innovation answered that it was unreasonable to require address holders to return and re-justify allocations whenever use changed, because ISPs regularly change customer assignments and service configurations. IGP's analysis concluded that AFRINIC may have had arguments under the contract, but widespread enforcement of such a theory would leave many members living under a permanent risk that ordinary business evolution could trigger review.
That is the institutional-economics problem. Discretionary review of use after allocation creates a hold-up right. Once a network has invested around a block, migrated customers, built reputation and integrated the addresses into systems, the registry's threat of adverse action becomes more powerful than it would have been before the investment. The holder cannot easily walk away. The customer cannot easily renumber. The buyer cannot assume a clean transfer. The registry may not need to extract money to impose cost; delay, uncertainty and the risk of non-recognition are enough.
The analogy to capital controls again helps. Investors fear not only confiscation but the administrative right to reopen past approvals. If a government can ask years later whether a factory still matches the original investment plan and can block dividend remittance or asset sale if the answer is unsatisfactory, investors price that risk from the start. Likewise, if a registry can ask whether an IPv4 block still matches original need, customer geography or business model, every buyer, lessee and lender discounts the block.
The legal aftermath also shows how registry overreach can damage the institution itself. IGP reported that Cloud Innovation obtained court relief, AFRINIC's bank accounts were provisionally frozen, and multiple cases accumulated in Mauritius. The Register later chronicled years of governance paralysis: AFRINIC operated without a board from 2022, entered receivership, attempted elections in 2025, saw one election suspended and annulled amid proxy-vote concerns, and only later elected a board under continuing legal and political scrutiny. One does not need to choose a hero in that saga to draw the lesson. A registry with high-consequence discretion and limited downside can provoke a legal arms race. The result may harm every member, including the members the registry says it is protecting.
The counterfactual is not laissez-faire chaos. AFRINIC had legitimate reasons to care about fraud, stolen blocks, duplicate claims, contact accuracy and routing security. A former insider's alleged manipulation of records, reported by MyBroadband and summarised in independent accounts, would make any registry more vigilant. But fraud control is not the same as commercial-use control. A forged company, stolen address block, false officer or duplicate claim justifies strong intervention. A debated theory about leasing, regional use or changed customer base requires a different standard: objective rules, proportional remedies, notice, cure, independent review and preservation of operational continuity while facts are tested.
Legitimacy depends on exit as much as voice
RIR governance traditionally relies on community participation: open mailing lists, public policy meetings, rough consensus, board ratification and member processes. AFRINIC's policy manual describes openness, transparency and fairness; anyone may participate in policy discussions. That model is often valuable for technical coordination. It helps reveal operational problems and creates norms across networks that must interoperate. But voice is not enough when the policy affects capital.
The reason is representation. A mailing-list participant is not necessarily authorised to bind an ISP's balance sheet. A consultant may speak without bearing operational downside. A registry employee may influence a process the registry later enforces. A government actor may care about industrial policy rather than holder reliance. A small operator may need liquidity while a larger incumbent may prefer restrictions that raise rivals' costs. A policy room can produce useful knowledge, but it cannot automatically become a legislature for every affected principal.
Exit is the missing discipline. In ordinary markets, bad governance is disciplined by the ability to leave, sell, switch providers, move assets, refinance or litigate under clear rights. RIRs are unusual because the registry layer is not easily substitutable. If AFRINIC controls the record for a resource, the holder cannot simply choose a competing African registry tomorrow. If transfer rules block outbound movement, even exit through sale is constrained. If provider-aggregatable space is non-portable, customer exit may require renumbering. If membership standing affects services, contractual dependence remains.
Lu Heng's public notes on portability and registry power make this point directly: networks need the ability to move resources away from a failing or hostile registry, not because portability is convenient, but because it disciplines governance. Whether one agrees with his preferred solutions or not, the economic principle is orthodox. Monopoly-like institutions require either strong public-law accountability or meaningful exit. A private-law registry that resists both becomes unstable once the resource it administers becomes valuable.
AFRINIC's recent governance history makes legitimacy more than an abstract concern. The Register reported in July 2025 that the receiver annulled an election after concerns about voter documentation and alleged powers of attorney; ICANN criticised the lack of transparent reporting and reserved rights under policy mechanisms for dysfunctional RIRs. The Register reported in September 2025 that AFRINIC finally elected eight directors, seven endorsed by Smart Africa, while critics still questioned arrangements, courts remained relevant and a criminal investigation into the earlier election was under way. In May 2026 The Register reported that ICANN intervened in a winding-up application and emphasised that numbering resources administered through AFRINIC are not assets available for distribution in liquidation.
Those facts do not decide the capital-control question by themselves. They do show why governance risk matters. Transfer restrictions might be less dangerous in an institution with stable leadership, uncontested elections, clear appeal paths, audited process and strong liability. In a registry emerging from receivership and litigation, discretionary controls over valuable assets carry a heavier legitimacy burden. The question for members is not only "did a policy pass?" It is "who bears the loss if this policy is wrong, implemented unevenly or used under institutional stress?"
Collateral-like reliance without admitted property
The RIR system has long resisted the idea that number resources are owned property. AFRINIC's manual uses public-resource and custodian language. The ASN section says neither assignment nor registration confers ownership and that users are custodians rather than owners. IPv6 policy principles say address space should not be considered freehold property and is licensed for use rather than owned. Official and quasi-official materials often maintain similar distinctions for IPv4, even while market practice treats IPv4 as valuable.
There are good reasons to avoid simplistic property language. IP addresses are globally unique identifiers whose usefulness depends on coordination. A holder cannot use a prefix if the rest of the routing system refuses to recognise it. Registry accuracy matters. Fraud, hijacking and duplicate claims can hurt third parties. Public-law property categories do not map neatly onto a global technical ledger administered by private regional entities.
Yet denying property language does not erase reliance. Businesses can have economically protectable reliance even where legal title is complex. A lease can be valuable without ownership. A licence can be transferable or non-transferable. A concession can be bankable if rules are stable. A domain name is not land, but a company may rely on it as a critical asset. Software licences, spectrum rights, airport slots, carbon credits and securities entitlements all show that markets can attach capital value to structured rights that are not simple freehold property.
IPv4 sits in that family. The holder relies on continued recognition and operational usability. Customers rely on continuity. Buyers rely on transfer recording. Lenders and investors rely on predictable treatment. Courts may be asked to protect the position when a registry threatens adverse action. If a registry says the holder does not own the address, that does not mean the registry may behave as owner. At most it means the relationship must be specified more carefully: what exactly may the holder do, what may the registry refuse, what process applies, what remedy exists, and how operational continuity is preserved during disputes?
Capital controls thrive in ambiguity. If the asset is valuable enough to control but not recognised enough to protect, the administrator gains leverage. It can say the holder has no property right when the holder seeks exit, and say the holder must obey resource conditions when the registry seeks control. It can treat the resource as not a commodity when a holder wants to sell, but as a scarce public input when the registry wants to ration. It can deny monetary valuation while imposing rules that change monetary value. This is the unstable middle ground.
A better registry model would be asset-neutral rather than asset-denying. It need not declare IPv4 to be freehold property. It should acknowledge that resource holders have reliance interests, transfer expectations and operational continuity needs that deserve procedural protection. It should separate registry functions from economic control. It should record rather than authorise ordinary transactions, reject updates only for enumerated objective reasons, and route ownership-like disputes to courts or independent adjudication rather than converting policy ambiguity into administrative punishment.
The counterargument deserves a serious answer
AFRINIC's defenders are not simply inventing concerns. The IPv4 market can invite hoarding, fraud, opacity, abuse and speculation. A region with fewer historical allocations worries that wealthy foreign networks will buy scarce resources. Registries face pressure when inaccurate records contribute to spam, malware, hijacking or law-enforcement frustration. Some operators may warehouse addresses while smaller networks struggle. Leasing structures can obscure who actually controls a prefix. If RPKI or reverse DNS is mismanaged, innocent users may be affected. A registry that does nothing would not be doing its job.
The answer is to distinguish technical invariants from economic preferences. Uniqueness is a technical invariant. Accurate records are a technical invariant. Fraud prevention in the registry record is a technical invariant. Contactability is a registry function. Security metadata, if offered, must be reliable. Dispute status should be visible enough to prevent false certainty. Operational continuity should be protected. These aims justify rules because they protect the ledger rather than a preferred business outcome.
Regional retention, anti-speculation morality, business-model approval, customer-geography review and indefinite needs reassessment are different. They are economic preferences. They may be motivated by development concerns, but they are not required for the Internet to route. They also create costs that fall on the very networks they claim to help. If policymakers want to subsidise African networks, they can consider grants, pooled procurement, transparent market-making, financing support, IPv6 operational assistance, tax treatment, public-sector demand aggregation or targeted support for IXPs and critical infrastructure. Using a private registry's ledger to suppress exit value is a blunt instrument.
The fraud concern is best answered by better title-like recording, not more discretion. Clear transfer records reduce fraud. Objective proof-of-control standards reduce fraud. Public conflict flags reduce fraud. Standardised escrow and settlement processes reduce fraud. Optional lessee contact fields reduce opacity. Defined cure periods reduce accidental defaults. Independent review reduces selective enforcement. By contrast, discretionary approval can increase fraud by making the official path slow, uncertain or politically risky. When formal channels are unsafe, markets route around them.
The routing-table concern is also real but limited. Portability and fragmentation can increase routing burden. But using registry non-portability as a provider lock-in tool is not a clean solution. Routing externalities should be managed through filtering norms, aggregation incentives, pricing of deaggregation where appropriate, operational coordination and transparency. Forced renumbering can impose private costs far larger than the public routing benefit, especially on enterprises with embedded systems, allowlists, compliance dependencies and customer contracts. The remedy must match the externality.
The public-interest argument is strongest for future free-pool allocations. If AFRINIC still issues scarce addresses from a remaining pool, it can impose forward-looking conditions: minimum and maximum sizes, documented need, critical-infrastructure priorities, anti-fraud checks and time-limited obligations. A recipient can decide whether to accept those conditions before receiving the resource. Applying new restrictions to existing resources, transfers, leases or imported blocks is more problematic because it changes reliance after investment.
What a narrower AFRINIC would still do
Reducing capital control does not mean abolishing the registry function. It means making the registry more precise. AFRINIC should know who is the recognised holder of a resource. It should maintain WHOIS and RDAP data. It should support reverse DNS, IRR and RPKI services. It should prevent duplicate registrations. It should correct stolen or fraudulently reassigned blocks. It should process transfers quickly when objective conditions are met. It should publish service levels. It should maintain an audit trail. It should flag disputes without destroying operational continuity. It should provide members with stable, predictable support.
The transfer test can be narrow. Is the transferor the recognised holder or legally authorised representative? Is the block uniquely identified? Is there an active court order, fraud hold, duplicate claim or independent adjudicative freeze? Has the transferee supplied accurate contact and organisation information? Can reverse DNS, RPKI and registry records be transitioned without breaking uniqueness or security integrity? If yes, the registry should record the transfer. It should not decide whether the buyer's business model is sufficiently African, whether the price is too high, whether the use is morally speculative, or whether the customer's customers sit in the approved geography.
Leasing can be treated similarly. The holder of record remains responsible for registry accuracy. The registry may offer fields for delegated operational contacts, abuse handling, route authorisation and time-limited arrangements. It may require that the holder remain reachable and that no duplicate claim be created. It may mark data quality. It may provide APIs. It should not convert leasing into presumptive non-compliance.
Membership good standing should be separated from existential resource harm wherever possible. Non-payment of fees may justify ordinary collection steps, late penalties, service limitations unrelated to correction, or suspension of discretionary services. Immediate revocation or transfer denial for curable administrative failures should be rare, proportionate and independently reviewable. A live network should not face a death penalty for a billing, contact or paperwork defect unless fraud, abandonment, duplicate claim or binding legal order is present.
Inbound resources require special care. If AFRINIC-region networks need IPv4 from the global market, the registry should make import safe. Imported blocks should not become trapped merely because they touched the AFRINIC system. Legacy or equivalent status should not be stripped by implication. Future outbound transfer should remain available. Registry service should be a service, not a conversion event. The easier it is for capital to leave, the more willing it is to enter.
Finally, policy scope needs a hard boundary. A proposal should identify the invariant it protects. If the invariant is uniqueness, accuracy, fraud prevention, contactability, security integrity, transfer recording or operational continuity, it belongs in the mandatory registry layer. If the aim is regional industrial policy, anti-speculation, price control, commercial morality or redistribution, it belongs elsewhere unless affected holders opt in or a competent public authority creates a lawful framework with due process and liability. A private registry should not smuggle economic governance into database policy.
Why this matters beyond AFRINIC
AFRINIC is the acute case because its crisis has been public, legal and prolonged. But the underlying problem is not African exceptionalism. All RIRs inherited an administrative model from a period when addresses were abundant or underpriced. All face the tension between community policy and asset value. All operate through private or nonprofit legal forms while administering a globally necessary coordination layer. All depend on member trust. All can be tempted to treat their service region as a source of legitimacy broader than registry service actually requires.
The difference is that AFRINIC's stress arrived first and noisily. The region had late allocations, scarcity politics, a high-profile address-market actor, a staff-theft scandal, court battles, receivership, election disputes and ICANN concern. That combination forced a question the rest of the system would prefer to answer slowly: what happens when the bookkeeper's rules can change the value of capital?
One answer is to defend the old language more aggressively. Declare resources public, region-bound and non-property. Treat market actors as threats. Tighten review. Add compliance duties. Insist that community process gives legitimacy. Seek political support when courts or members challenge the institution. This path may feel stabilising in the short term. It concentrates power precisely when trust is weakest.
The better answer is institutional humility. A registry's authority is strongest when it does less and does it predictably. It should be boring infrastructure. It should make the ledger more accurate than the grey market, the official transfer path safer than private improvisation, and the dispute process calmer than emergency litigation. It should not try to be the industrial-planning ministry of IPv4, the customs authority for routes, the moral regulator of leasing, or the owner that refuses to say it owns.
The stakes are practical. If AFRINIC-style controls spread, IPv4 markets will not disappear. They will become more complex, more legalistic and less transparent. Operators will build wrappers around restrictions. More value will move through leases, service contracts and offshore structures. More disputes will reach courts. More holders will price registry risk into every deal. More customers will discover that their network identity depends on an institutional layer they barely understood. The ledger will lose authority because it asked for too much.
If, instead, AFRINIC narrows its role, it could turn crisis into a useful precedent. It could show that an RIR can preserve uniqueness while respecting market reality; that African networks are better served by liquidity than by confinement; that member governance is stronger when exit exists; and that scarce digital infrastructure needs predictable recording more than moralised control. The result would not be a libertarian fantasy. It would be a disciplined registry.
Uncertainty and watchpoints
Several uncertainties should temper any confident conclusion. The legal merits of specific AFRINIC disputes remain fact-dependent, and public reporting does not substitute for court findings. Not every transfer restriction has the same effect; implementation details, appeal rights and service timelines matter. IPv6 deployment may reduce some future demand for IPv4, though dual-stack reality keeps IPv4 economically relevant for now. Some African operators may prefer regional retention if they fear being priced out by global buyers. Fraud and stolen-address histories justify stronger verification than a pure market model would allow. ICANN's evolving approach to failing or unstable RIRs may change incentives before AFRINIC settles into a durable pattern.
Start with the final implemented transfer framework. The key issue is not whether AFRINIC has a transfer policy, but whether it treats transfers as economic transactions requiring discretionary permission or as record updates requiring objective verification. Service-level deadlines, written reasons for refusal and appeal paths will reveal the real posture.
Outbound mobility is the price signal. If AFRINIC-issued resources cannot leave the region on the same practical terms as comparable resources elsewhere, a regional discount is likely. Evidence would appear in transaction spreads, seller reluctance, inbound-transfer hesitation, longer deal timelines and a preference for leasing or nominee structures over clean sale.
Inbound status is the supply signal. Imported or legacy resources should not lose economically important status merely by entering the AFRINIC registry. Any policy that converts imported capital into a trapped regional object will discourage the supply Africa needs.
Leasing recognition will separate recordkeeping from control. A serious registry will seek better visibility into leasing, delegation and customer contact arrangements. A control-oriented registry will treat leasing as a violation signal. The former improves data quality; the latter pushes activity into shadows.
Membership and fee enforcement deserve the same scrutiny. Billing discipline is ordinary. Using good-standing issues to block transfers, threaten resource continuity or create leverage over unrelated disputes can turn routine administration into capital control. Closure-process rules should be watched for proportionality and cure rights.
The use of RPKI, IRR, reverse DNS and abuse-contact mechanisms is the operational test. These tools are valuable because they support security and contactability. They become dangerous if converted into enforcement weapons for commercial or regional policy. A failed mailbox should not become a route-security death sentence.
Governance legitimacy remains the institutional test. AFRINIC's elected board, receivership legacy, pending litigation, ICANN interventions, Smart Africa influence and member-rights debates all affect confidence. Policies that might be tolerable in a settled institution may be destabilising in one still proving its mandate.
The last watchpoint is market adaptation. If operators begin treating AFRINIC records as less reliable, if more contracts route around registry transfer channels, or if customers prefer first-party lessors specifically to avoid registry exposure, the system will have received its verdict. Capital controls often look effective at the administrative window before they show their cost in discounts, avoidance and informal markets. AFRINIC's challenge is to avoid learning that lesson at the registry layer.