Conservation after the free pool
AFRINIC is usually described in the calm dialect of internet administration. It is a regional internet registry. It serves Africa and parts of the Indian Ocean. It records holders of internet number resources, distributes IPv4 and IPv6 addresses and autonomous system numbers, and operates adjacent services such as WHOIS, RDAP, reverse DNS, routing-registry functions and RPKI. Those facts are useful. They are not enough. The argument about AFRINIC's conservation language is not settled by reciting the registry's service catalogue. It turns on what that language does after IPv4 scarcity has become an economic fact.
Conservation once had a limited and intelligible meaning. When a registry still held a pool of unallocated IPv4 addresses, it needed rules for who should receive them and how much should be issued at a time. AFRINIC's Consolidated Policy Manual is useful here as a factual exhibit. It lists uniqueness, registration, aggregation and conservation as goals of the registry system. It says public unicast IPv4 addresses must be globally unique. It says allocations and assignments should be registered. It says aggregation helps routing. It then defines conservation around actual need and immediate use, while discouraging stockpiling and reservations.
In the allocation era that was not an absurd doctrine. A registry giving out a scarce but administratively priced resource had to avoid waste. Need tests, immediate-use tests and anti-stockpiling rules could be defended as rationing devices. They reduced the chance that the best-informed or best-staffed applicants would drain the common pool before others could make a plausible case. In that world conservation looked like technocratic restraint. The registry was not yet supervising a mature market. It was rationing a pool.
The trouble begins when the same vocabulary follows the resource into a different economy. IPv4 addresses are no longer merely pending allocations from a common administrative stock. They are embedded in networks, customer relationships, leasing contracts, hosting businesses, security reputations, allowlists, reverse DNS records, route filters, financing assumptions and corporate valuations. A block that once looked like a database entry now carries cash-flow consequences. When a registry invokes conservation over such a block, it is no longer only preventing waste. It is influencing asset movement.
Lu Heng's public note "The Policy Mirror", published on 30 June 2026, gives a direct statement of that transition. Its useful claim is not that a registry has no function. It is that the function is narrow: preserve uniqueness, keep accurate records, record transfers, support contactability, maintain security metadata and isolate disputes. Once the registry begins to decide where value may move, whether a business model is proper, which region may receive a transfer or whether a holder's evolving customers still satisfy an old justification, the ledger has become a gate.
Those public notes should be read with attribution. Lu is a participant in this market and has direct interests in the controversies around AFRINIC, Cloud Innovation and IPv4 leasing. His claims are not court findings. But interested participants can still identify an economic mechanism that institutional language obscures. The structural point is this: conservation rhetoric can hide distribution choices. It can make a decision about who controls scarce capital sound like a decision about technical hygiene.
AFRINIC's own exhaustion material shows why the change matters. The registry says it entered IPv4 Exhaustion Soft Landing Phase 2 on 13 January 2020. In that phase the minimum IPv4 allocation or assignment is /24 and the maximum is /22. Requests are processed through tickets, evaluated against policy and contractual checks, and additional requests require efficient use of previously delegated space. Whatever one thinks of that design, it is not an engine of abundance. It is late-stage rationing for a nearly depleted pool.
Once the free pool can no longer meet regional growth, conservation changes character. The central policy question is no longer how to distribute newly issued addresses cheaply. It is how existing addresses move from lower-valued use to higher-valued use, how underused blocks become visible, how African operators obtain capacity, how imported resources avoid becoming trapped, and how customers maintain continuity while IPv6 remains incomplete. Conservation rhetoric answers too many of those questions by pretending they are still one question: who deserves permission from the registry?
That is not a technical answer. It is an allocation philosophy.
Soft Landing as rationing, not salvation
The phrase "Soft Landing" has an attractive administrative calm. It suggests an orderly descent, a gradual transition, a pool managed carefully enough to avoid shock while networks prepare for IPv6. The economics are harsher. Soft Landing is a rationing system for the end of the free pool. It cannot supply the future address demand of African networks. Nor can it solve the mismatch between administratively priced allocations and market-priced IPv4.
AFRINIC's exhaustion page records the sequence. The global IANA pool reached its final-allocation moment in 2011. Other RIR regions exhausted their free pools earlier. AFRINIC entered Phase 1 of its Soft Landing process in March 2017 and Phase 2 in January 2020. Under the public explanation, Phase 2 begins when only a defined amount of non-reserved space remains in the final /8. In Phase 2, the maximum request is /22. A /22 is 1,024 IPv4 addresses. That may be useful for continuity in a limited deployment. It is not enough to build a continent-scale internet economy.
Soft Landing is therefore better read as a signal than as a solution. It tells operators that the old model has ended. It tells them that large increments are no longer available by ordinary application. It tells them that documentation, contractual checks, first-come processing and efficient-use thresholds are scarcity procedures. Such rules may preserve orderly distribution at the margin. They do not answer where the next large pools of working IPv4 will come from.
The Internet Governance Project's 2021 analysis of the AFRINIC crisis used the phrase "a fight over crumbs". It argued that the future growth of Africa's internet could not be sustained by the small leftover portion of IPv4 space controlled by AFRINIC. Growth would require importing large numbers of IPv4 addresses from the market, relying more heavily on IPv6, or both. The language was sharp, but the arithmetic is difficult to escape. If the remaining pool is small, the economic role of policy shifts from allocation to mobility.
Mobility is where conservation rhetoric becomes dangerous. A registry may say it is conserving addresses for the region. In practice it may be reducing the liquidity of addresses already held by regional operators. It may also discourage outside holders from moving resources into the region if entry creates status loss, exit restrictions or continuing exposure to policy reinterpretation. A wall around a small pool does not create more water. It can make everyone inside the wall poorer and thirstier.
The old conservation model also confuses documented need with social value. A large, well-lawyered, well-staffed operator is usually better at proving need than a small one. It has engineers who can produce plans, lawyers who understand contracts, finance staff who manage invoices and executives who know how to navigate institutional procedures. A small ISP in a thin market may have a real commercial need and a weak administrative file. A needs-based system is apt to read those differences as if they were differences in merit.
That is one of Lu's recurring points in the "Poverty Penalty" note. The RIR system was built to coordinate globally unique number resources and administer scarcity through registration, documentation, utilisation review and policy. It was not built as a global redistribution mechanism. If more customers, devices, staff and deployed capital create more documented need, richer and larger networks will predictably fare better inside the process. The process does not stand above inequality. It reads inequality and gives it administrative form.
Soft Landing therefore creates an awkward political economy. The registry's moral language suggests fairness, caution and regional responsibility. The operational reality is a small remaining pool, a thickening permission process and an unavoidable secondary market. A region that needs future capacity must care about transferability, leasing, inbound status, registry reliability and price discovery. Those are market institutions. They are not repaired by pretending conservation can still do the work of abundance.
The point is not that Soft Landing was foolish when adopted. It is that a policy built for exhaustion should not become a permanent claim of registry authority over every later economic use of IPv4. A rationing rule for residual free-pool issuance is one thing. A continuing veto over market movement is another. The former conserves a pool. The latter controls capital.
The asset nobody wants to name
The RIR tradition resists property language, often for defensible reasons. An IP address is not land. It is not a machine. Its usefulness depends on global uniqueness, routing acceptance, registry records, security assertions and operational coordination by many parties. A holder cannot make the rest of the internet recognise a prefix merely by invoking ownership. The resource is not an ordinary private thing severed from a coordination system.
Avoiding simplistic property talk, however, is not the same as denying economic reality. IPv4 has become scarce, transferable, priced, leased and litigated over. It supports revenue. It can be valued in business acquisitions. It can be rented to customers. It can be treated as a continuity input by hosting companies, security vendors, SaaS platforms, access networks and enterprises whose systems still depend on IPv4 reachability. A company may not own addresses as it owns furniture, but it can suffer real loss if recognised use, transferability or routing credibility is impaired.
Lu's note on the present registry model becoming impossible once IPv4 becomes a real asset states the contradiction clearly. The legal and institutional shell was built for a low-value administrative record; the economic substance underneath has become strategic. Registry contracts and policy structures still often look like service arrangements for a database. Downstream reliance can be orders of magnitude larger than the fees paid to the registry.
AFRINIC's fee schedule illustrates the gap without needing ideology. Annual membership fees are based on categories derived from resource holdings. A medium LIR category covering /16 to less than /14 carries a membership fee far below the market value of a /16 in recent transfer-market estimates. The Internet Governance Project noted in 2021 that IPv4 prices had moved from roughly $8 per address in 2017 to about $30 per address by 2021, making a /16 worth roughly $2m at that time. AFRINIC's fee is not the asset price. It is the service charge for the institutional relationship.
That gap is not automatically scandalous. Registries are not supposed to auction every resource to the highest bidder. But the gap explains why conservation language becomes combustible. If a resource that costs thousands of dollars a year to maintain in registry fees can support millions of dollars of market value, every discretionary interpretation of policy becomes economically meaningful. Needs review is not paperwork. Transfer approval is not clerical. Good-standing status is not just billing. Legacy-status conversion is not merely semantic. Each changes the expected value of the resource.
Assetisation also changes the burden of uncertainty. In an abundant world, a registry delay is irritating. In a scarce world, it can be a financing cost, a lost customer, a failed transaction or a litigation trigger. A buyer needs to know whether a block can be transferred. A seller needs to know whether proceeds can be realised. A lessee needs continuity. A lender needs to know whether a business dependent on addresses has a stable input. An operator needs to know whether changing customers, routing patterns or geography creates compliance risk.
Public LARUS materials, again as interested market evidence rather than neutral proof, sell first-party IPv4 leasing partly as a way to place registry-layer exposure upstream from the customer's operating company. The significance is not the marketing claim itself. It is the fact that continuity and registry risk are now commercial product features. A market does not invent that language unless customers recognise the risk it names.
This is where conservation rhetoric can become a disguise. By insisting that IPv4 is not property, an institution may try to avoid the duties associated with impairing valuable reliance. By insisting that it is conserving a public resource, it may justify restrictions that change private balance sheets. By insisting that the registry is merely enforcing policy, it may understate the fact that policy choices decide whether capital can move.
A more honest vocabulary would be asset-neutral. It need not say IPv4 is freehold property everywhere and for all purposes. It should say that registry-recognised number resources create reliance, transfer expectations and operational dependence that deserve predictable treatment. The registry may record and coordinate. It should not deny economic value in one paragraph and control economic movement in the next.
That denial is the heart of conservation rhetoric after scarcity. The asset is said not to be an asset when the holder asks for market freedom. It becomes a scarce public resource again when the registry asks for control.
The poor-country argument and the arithmetic of control
The strongest defence of AFRINIC-style conservation is moral rather than technical. Africa received relatively little IPv4 address space compared with earlier and richer regions. African networks still need capacity. If addresses can be freely sold, leased or moved outside the region, richer buyers may drain the scarce pool and leave poorer operators paying global prices for a resource that should have supported local development. The argument is intuitive. It is also incomplete.
The first problem is historical. The RIR system did not allocate IPv4 according to poverty. It allocated according to documented network need, timing, administrative capability and the growth patterns of the internet economy. Lu's "Poverty Penalty" note cites distribution numbers to make the point: the United States and China together hold more than half of allocated IPv4 in one referenced distribution report, while AFRINIC's regional share is a small fraction of the global delegated total. Inside Africa, the same note says South Africa, Egypt and Morocco together hold a large share of the sovereign-state total. Whether one disputes a particular dataset or date, the pattern is not egalitarian.
The second problem is mechanical. A regional transfer wall does not create more IPv4. It changes the set of buyers, sellers and intermediaries who can transact through the official path. A narrower market usually lowers the exit value of the asset. Lower exit value reduces collateral usefulness. It reduces the incentive to discover underused supply. It encourages long-term leases, nominee structures, customer delegations and informal arrangements that may not be fully visible in registry data. The official rhetoric says the resource is being conserved. The practical effect may be opacity.
The third problem is distributional. Rich operators handle discretion better than poor operators. A large multinational can hire counsel, structure around policy, wait through review, maintain multiple regional relationships and buy addresses elsewhere. A small regional ISP cannot. If the official transfer process is slow, ambiguous or politically risky, the small operator pays a higher proportionate cost. If the market is transparent, the same operator still faces price, but price can be compared, budgeted, financed, negotiated or avoided through leasing. Discretion is harder to finance.
That is why the claim that commercialisation is inherently anti-poor should be treated with caution. Commercialisation does not guarantee justice. Rich buyers can buy more. Speculative behaviour can distort supply. Fraud and abuse can exploit transition windows. But the alternative is not a benevolent administrative paradise. It is gatekeeping by institutions whose processes already tend to reward incumbency, paperwork capacity and insider familiarity. The serious question is which system gives weak operators lower total cost: visible price plus clean rules, or lower nominal price plus delay, uncertainty and dependence on discretion.
The conservation story also has a convenient villain problem. Public debate often personalises the issue around Lu Heng, Cloud Innovation, LARUS or the Number Resource Society. Some of those actors have strong commercial interests; some claims have been contested; courts and regulators have not resolved every allegation. Readers should not treat participant assertions as judgments. Yet focusing too much on one actor can obscure scale. A few million or even ten million addresses cannot explain the global distribution of IPv4, the underuse of legacy allocations, or the fact that poorer networks arrived late to an already concentrated resource landscape.
KrebsOnSecurity's 2019 reporting on alleged AFRINIC address heists is important for a different reason. It showed that underpriced, scarce IPv4 and weak registry controls can invite fraud, manipulation and shadow markets. That is a real concern. But fraud control is not the same as regional capital control. A registry can require proof of authority, verify records, protect against forged transfers, publish dispute status and preserve audit trails. Those are ledger functions. They do not require the registry to decide whether every buyer, lessee or customer geography satisfies a development narrative.
The poor-country argument therefore needs a more precise policy target. If the goal is affordable African connectivity, the instruments are power, backhaul, fibre, spectrum, competition, neutral IXPs, data centres, financing, local skills, procurement, customer income and eventual IPv6 capability. IPv4 matters because compatibility still matters. But suppressing the liquidity of IPv4 assets held by African operators is a poor substitute for industrial policy. It gives the appearance of solidarity while reducing the financial usefulness of the one scarce resource some operators actually hold.
A genuinely pro-poor registry would make access cheaper in total, not merely cheaper in rhetoric. It would reduce transaction friction. It would make transfers predictable. It would make leasing visible rather than shameful. It would protect inbound status. It would lower fixed compliance costs. It would separate fraud review from business-model review. It would not ask the weakest networks to accept administrative dependence as proof that someone is looking after them.
Regional retention as capital control
Regional retention sounds like common sense until it is translated into asset economics. If a resource is labelled African because it was issued through AFRINIC, and if that label restricts transfer out of the region, the label changes value. It is no longer just metadata. It is a capital-control device. It defines the permitted market for a scarce input.
The phrase capital control is an analogy about mechanism, not an accusation of identical legal form. AFRINIC is not a central bank and IPv4 is not a currency. But capital controls limit convertibility, exit, sale, remittance or movement of assets across borders. Their effect is often to preserve an official story while creating a discount. A bank balance trapped behind exchange restrictions is still a bank balance, but it is not worth the same as freely transferable cash. An IPv4 block trapped behind regional transfer restrictions may still route, but it is not worth the same as a block that can be sold or moved freely.
Lu's "Policy Mirror" note argues that AFRINIC's 2026 transfer architecture classifies resources by origin and status, restricts some resources from leaving the AFRINIC region, requires written approval for transfers and brings incoming resources under AFRINIC's policy environment. That is Lu's public reading of the policy framework, not an independent court finding. It does, however, match a broader problem already visible in the older Consolidated Policy Manual: transfers within the AFRINIC region require the recipient's need to be approved by AFRINIC; the recipient must be an AFRINIC member subject to current policies and must sign the Registration Services Agreement; and transferred legacy resources no longer retain legacy status under that older section.
Those clauses matter because they make the registry a gate for movement. The source must be recognised and undisputed. The recipient must justify need. The recipient must enter the membership and contract structure. Status can change after transfer. Good-standing requirements in the fee schedule also affect whether transfers are considered. None of these rules is dramatic alone. Together they produce a permission architecture over a scarce, priced resource.
Regional controls also create a supply paradox. If African networks need more IPv4 than AFRINIC's remaining pool can supply, the region benefits from attracting addresses from elsewhere. But capital that enters a region wants to know whether it can later leave. If inbound resources lose valuable status, become subject to broader policy or face future outbound restrictions, sellers and buyers will demand a discount. Some will avoid the region entirely. A policy meant to protect local access can therefore reduce the inbound supply that local access requires.
This is an old development-policy error. A government or institution sees an asset leaving and assumes exit is the loss. It restricts exit. The restriction then reduces the asset's price, reduces investment, discourages future entry and encourages informal workarounds. What looks like protection becomes a tax on the protected. The proper comparison is not between resources leaving and resources staying. It is between a liquid asset market that can finance regional operators and an illiquid administrative market that locks them into lower value.
There is also an information cost. When official transfers are burdened, activity migrates to contracts that do not require a clean ownership-style transfer: leases, customer delegations, route sponsorship, long-term service arrangements, nominees or offshore structures. Some of these arrangements are lawful and useful. Some may be risky. The registry's interest should be to make them legible where possible. A policy that condemns them as leakage may reduce visibility and then cite reduced visibility as a reason for more control.
The route itself does not care about region. A block can be registered in one service region, routed through another, used by customers in a third and monetised by a company incorporated in a fourth. That does not mean geography never matters. Corporate law, sanctions, tax, consumer protection and network licensing all have geographic components. But those are not reasons for a private registry to treat a service-region label as title. Geography can be operational metadata. It should not be a capital cage.
The essential distinction is simple. A registry may say: we need accurate records of who controls the resource, how to contact them, whether a dispute exists and what security assertions are valid. That is conservation of the ledger. It may not plausibly say: because we serve a region, we may decide where the economic value of a globally routed identifier may move. That is conservation rhetoric turned into capital policy.
Leasing, dependency and the operational user
IPv4 leasing is often treated as a loophole in conservation debates. That is the wrong starting point. Leasing is a predictable response to scarcity, price, operational risk and uncertain direct holding. A company may need address capacity for a service without wanting to buy a block. A small operator may prefer operating expenditure to capital expenditure. A hosting company may need flexible pools for customer churn. A security vendor may need dedicated space for reputation management. A multinational may need continuity across jurisdictions. These are ordinary commercial motives, not proof of abuse.
The older assignment language in AFRINIC's manual says assignments are for specific purposes documented by specific organisations and are not to be sub-assigned to other parties. That makes sense if the registry is trying to ensure that an address block issued from a pool is used for the stated deployment rather than resold immediately. It is harder to sustain once the market is mature and operational delegation is routine. Cloud services, hosting, broadband, enterprise networking and managed security all involve layers of use that do not map neatly onto the original applicant's declared purpose.
If a registry treats leasing as presumptive evasion, it creates bad incentives. Holders will keep public records thin. Lessees will rely on private contracts. Abuse contacts may not reflect the party best positioned to act. Route-authorisation relationships may be harder to interpret. Customers may become dependent on arrangements that exist in the network but not in the registry. The database becomes less accurate because the official story is safer than the full story.
A better registry would separate recognised holder from operational user. The holder of record can remain responsible for uniqueness and registry-facing obligations. The operational user, lessee or customer can be recorded voluntarily or where legally required for contactability, abuse handling, routing authorisation, reverse DNS or time-limited control. Dispute status can be marked. Contractual disagreements can be isolated. The point is not to bless every lease. It is to make reality more legible.
Lu's notes on LARUS One and registry-layer risk present leasing as a way to move direct RIR exposure away from the customer's operating entity. The argument is interested, and readers should treat it as such. But it identifies a real buyer concern. Direct holding does not only give a company a clean name in the database. It also gives the company direct exposure to registry contracts, policy changes, reviews, invoices, dispute processes and possible adverse action. A first-party lessor claims to absorb some of that exposure upstream. Whether a particular firm delivers that promise is a factual question. The existence of demand for such structures is a market signal.
Operator dependency is the reason leasing should not be dismissed. A customer using leased addresses may still build real dependence: firewall rules, reputation, mail configuration, VPN access lists, customer onboarding, payment systems, geolocation and contractual service commitments. If the registry destabilises the holder, the downstream user may suffer without ever having participated in the registry process. Conservation rhetoric aimed at the holder can therefore produce harm far beyond the holder.
This is one of the blind spots of member-centred governance. The registry sees a member and a resource. The market sees a chain of dependencies: holder, lessor, lessee, downstream customer, end user, transit carrier, cloud platform, security vendor and public service. A policy framed as discipline against one member can ripple through many operational relationships. That does not mean the registry must accept every commercial arrangement. It does mean remedies must account for customer continuity and collateral damage.
The temporary-resource section of AFRINIC's manual is oddly revealing. It explicitly contemplates resources assigned on a lease basis for short-term activities such as conferences or exhibitions. That is not the same as commercial IPv4 leasing, but it shows the concept of time-limited use is not alien to registry administration. The challenge is to design permanent registry records that reflect modern delegation without converting every delegation into a transfer or a violation.
Leasing can create abuse risks. A lessee may be hard to identify. Bad actors can burn address reputation and leave the holder exposed. Rapid reallocation can create blacklist lag or attribution error. Those problems justify better data, deposits, contractual controls, verified contacts and routing-security hygiene. They do not justify pretending leasing is not part of the scarcity economy. Conservation that refuses to see leasing conserves only the registry's preferred fiction.
Discretion is a liquidity cost
Liquidity is not just the presence of buyers and sellers. It is confidence that a transaction can close on known terms within a reasonable time and that the resulting position will be recognised. In ordinary financial markets, liquidity depends on settlement systems, custody, clear rules, reliable intermediaries and enforceable rights. In the IPv4 economy, the registry record plays a comparable coordination role. If the record is uncertain, the market becomes more expensive.
Registry discretion raises liquidity costs in several ways. A needs test after purchase delays closing and adds uncertainty. A regional transfer rule narrows the buyer pool. Good-standing requirements can turn unrelated billing or membership issues into transaction blockers. Legacy-status conversion changes the future value of a block. Ambiguous review rights make historical use relevant long after the original allocation. Broad revocation language makes every commercial plan hostage to later interpretation. Even if adverse action is rare, the possibility is priced.
The "enforcement creep" theme in Lu's public archive describes this as the replacement of price discovery by permission-seeking. The phrase is polemical, but the mechanism is familiar to economists. If a scarce asset cannot move without administrative approval, the approval process becomes part of the asset's price. Buyers discount. Sellers hesitate. Brokers earn more for navigating process than for matching supply and demand. Lawyers and consultants gain importance. Informal substitutes appear. The asset exists, but its velocity falls.
The harm is not limited to speculators. A small ISP sitting on underused space may be more willing to sell part of it if the market is reliable and the proceeds can finance fibre, power backup, security upgrades or customer acquisition. If transfer is uncertain or politically fraught, the operator may sit on the block. Conservation rhetoric then produces waste: the resource remains underused because the official path to productive reuse is costly.
Need-based transfer review is especially awkward in a market. The buyer's willingness to pay is evidence of expected use, though not perfect evidence. The buyer may be wrong, speculative or strategic. But the registry reviewer is not necessarily better at forecasting demand. The reviewer does not bear the buyer's capital cost, customer risk or opportunity cost. When the reviewer has veto power, the decision shifts from commercial risk-taking to administrative permission. The result is neither pure market nor pure planning. It is a hybrid that can inherit the worst of both.
The same logic applies to conservation of prices. Institutions may dislike rising IPv4 prices because they make scarcity visible. But suppressing price signals does not eliminate scarcity. It hides it in queues, favours, reviews, disputes, leases and compliance costs. A visible price may be uncomfortable. Hidden cost is often worse because it is harder to compare and easier for incumbents to absorb.
Registration accuracy suffers too. The more dangerous the registry becomes, the less candid the market will be. If a holder believes disclosure of leasing, customer geography or commercial delegation may trigger review, it will disclose less. If a buyer believes a transfer will be denied for a reason unrelated to fraud or control, it may structure around the transfer. If a lessee believes its use will be politicised, it may prefer private paperwork. The ledger becomes less truthful because truth is risky.
That outcome defeats the defensible part of conservation. The registry's public purpose is strongest when it keeps the authoritative record more accurate than the grey market. It should make the official path safer than improvisation. When it turns the official path into a discretionary tribunal, it invites the improvisation it later condemns.
There is a narrower model. Transfers could be recorded when the transferor is the recognised holder or authorised representative, the resource is clearly identified, no active fraud hold or court order prevents the update, the transferee provides accurate information, and reverse DNS and security records can be preserved or transitioned. Rejections would be written, evidence-based and appealable. That model conserves uniqueness and record integrity. It does not ask the registry to decide whether the transaction is morally regional enough.
Liquidity is not an indulgence for traders. In a scarce infrastructure input, liquidity is how idle or underused capacity becomes available to those who can use it. Conservation that destroys liquidity may conserve an institutional role while wasting the resource.
Governance turmoil and the scarcity discount
The economics of conservation rhetoric cannot be separated from AFRINIC's governance history. Discretion is less dangerous in a stable institution with transparent elections, clear appeal paths, reliable finances, predictable courts, audited processes and strong internal controls. It is more dangerous in an institution emerging from litigation, receivership, election disputes and allegations of record-integrity failure. The same rule has a different cost depending on who applies it.
KrebsOnSecurity reported in 2019 on allegations that a senior AFRINIC figure had been linked to companies selling valuable address blocks, with researcher Ron Guilmette estimating the market value of disputed addresses at more than $50m. AFRINIC's then new CEO said the organisation was investigating. Those allegations should not be inflated beyond the public record, but they matter as background. A registry invoking conservation after a record-integrity scandal must show that the cure is precise rather than opportunistic.
The Cloud Innovation dispute then turned registry discretion into a public stress test. The Internet Governance Project's 2021 account described AFRINIC as asserting concerns about out-of-region use, discrepancies between registered usage and actual use, and the need for changed or continuing justification. Cloud Innovation disputed those claims and argued that constant re-justification would make the registry a central planner for network operations. IGP criticised both sides: AFRINIC for overreaction and bad policy premises, Cloud Innovation for excessive legal tactics. That balance is useful because it avoids turning the dispute into simple hero and villain.
The legal aftermath was severe. Public reporting and NRO statements recorded receivership and efforts to reconstitute AFRINIC's board. The Register reported in July 2025 that an election had been annulled after concerns about voter documentation and alleged powers of attorney, with ICANN complaining about lack of transparent reporting. In September 2025 The Register reported that AFRINIC had elected eight directors, seven endorsed by Smart Africa, but remained exposed to criticism, court challenges and investigations. In May 2026 it reported that ICANN had intervened in a winding-up application, emphasising that numbering resources administered through AFRINIC were not assets available for distribution in liquidation.
None of that proves that any particular AFRINIC policy position is unlawful. It does prove that institutional risk is not theoretical. A buyer, seller, lessor, customer or lender looking at AFRINIC-linked resources must consider more than address scarcity. It must consider whether the registry can process requests, whether elections are accepted, whether courts may constrain action, whether ICANN or peer registries may intervene, whether bylaws are coherent under Mauritian law, and whether policies will be applied predictably.
That is the scarcity discount. IPv4 is already finite. On top of that comes registry risk. If the registry is seen as a neutral ledger, the discount is lower. If it is seen as a discretionary gatekeeper in a contested governance environment, the discount rises. The discount appears in lower bids, higher legal costs, greater preference for leasing, longer deal timelines, more warranties, more indemnities and reluctance to import resources into the registry's reach.
The conservation narrative may unintentionally worsen that discount. When an institution under stress insists that it must preserve broad authority for the sake of the region, it asks the market to trust its discretion precisely when discretion is most suspect. A narrower posture would be cheaper. It would say: whatever the governance dispute, the registry will keep records accurate, transfers objective, security services neutral, disputes isolated and running networks protected. That would reduce risk without requiring everyone to accept the institution's larger self-description.
Lu's "Registry Continuity Fallacy" note makes that distinction between the function and the gatekeeper. The function is real: uniqueness, accurate records, RDAP, WHOIS, reverse DNS, RPKI, running-network continuity and independent dispute handling. The fallacy is treating continuity of those functions as requiring preservation of every authority claimed by the current institution. A registry can be important without being sovereign. In fact, the more important the function, the more replaceable and auditable the operator should be.
For AFRINIC, governance repair and policy repair are therefore linked. A stable board and clean accounts would help, but they would not answer the asset question. Conversely, a narrow, objective transfer and recordkeeping posture would help restore trust even before every political wound has healed. Conservation rhetoric that preserves discretion is costly because AFRINIC is not operating from a position of unquestioned legitimacy.
What genuine conservation would conserve
The problem with conservation is not the word. It is the object being conserved. A serious registry should conserve uniqueness. It should conserve record accuracy. It should conserve contactability. It should conserve routing-adjacent security. It should conserve operational continuity during disputes. It should conserve audit trails and historical truth. These are the things a registry is institutionally suited to protect.
It should not conserve administrative underpricing after the pool is gone. It should not conserve regional immobility. It should not conserve the moral authority of a policy room over absent principals. It should not conserve an anti-market story that hides costs in procedure. It should not conserve the registry's ability to change the value of resources while disclaiming responsibility for economic consequences. Those are not technical invariants. They are distributional choices.
The first principle of a narrower conservation policy would be separation between free-pool allocation and post-allocation movement. If AFRINIC still issues residual resources from a small pool, it can apply documented-need and anti-fraud criteria prospectively. Applicants know the conditions before they accept the resource. But already held resources, transferred resources, leased resources and imported resources should not be governed as if they were still awaiting initial rationing. Movement should be denied only for objective defects: lack of authority, duplicate claims, fraud holds, binding orders, inaccurate records or inability to preserve essential registry continuity.
The second principle would be anti-retroactivity. Operators invest under assumptions. Customers build dependencies. Lessees and lessors structure contracts. A policy adopted later should not reduce transferability, portability, leasing ability or recognised control except for narrow technical or legal reasons. Conservation rhetoric often hides retroactivity by presenting new restrictions as clarification of old stewardship. Markets experience that as confiscation risk.
The third principle would be leasing recognition. A registry should prefer disclosed operational delegation to invisible operational delegation. It can retain the holder of record, but offer fields for delegated contacts, time-limited arrangements, routing authorisations and dispute notes. It can separate the question "who is recognised by the registry?" from the question "who is operating or using this block under contract?" That improves data quality and reduces collateral damage.
The fourth principle would be inbound safety. If a resource moves into the AFRINIC system, the holder should not fear that it has entered a trap. Legacy or equivalent status should not be stripped by implication. Outbound freedom should remain available. Registry services should be services, not conversion events. A region that needs imported IPv4 should make entry boring and exit credible.
The fifth principle would be security-service neutrality. RPKI, reverse DNS and related services should track recognised control and technical validity, not obedience in unrelated policy disputes. A broken abuse mailbox, disagreement over leasing or dispute about customer geography should not become a route-security weapon. Security mechanisms lose legitimacy if market actors perceive them as enforcement levers.
The sixth principle would be dispute isolation. When claims conflict, the registry should preserve the last verified operational state, mark the dispute, accept evidence and defer contested legal questions to an independent forum or court. It should not convert every ambiguity into revocation, reclaim or silent service impairment. The goal is to keep the network running while the claim is decided.
The seventh principle would be liability-power symmetry. If the registry wants broad discretion over transfers, usage and continuity, it should accept responsibility proportionate to the foreseeable harm of wrongful action. If it does not want that liability, it should narrow its discretion. The current model often wants the influence of an economic regulator with the risk profile of a membership service operator. That combination is unstable once the resource is valuable.
None of these principles abolishes registry governance. They make it more exact. AFRINIC would still record holders, publish contacts, verify authority, protect against fraud, manage reverse DNS and RPKI, maintain audit logs, process transfers, support IPv6 and publish statistics. It would still matter. It would simply stop using conservation as a basis for deciding how capital should move.
That is the institutional bargain a scarce IPv4 world requires. The registry gets legitimacy by doing less and doing it predictably. Operators get confidence that their reliance will not be reinterpreted away. African networks get a better chance of importing, financing, leasing and redeploying the resources they need. Conservation becomes technical again rather than moral theatre.
The watchpoints that will show the real policy
The first watchpoint is the final transfer framework as implemented, not as described in communiques. The key question is whether AFRINIC treats transfer as objective recording or discretionary permission. Written approval, needs tests, regional-use review, legacy-status loss and open-ended compliance checks point toward capital control. Proof of holder authority, fraud review, accurate transferee records, clear deadlines, written reasons and independent appeal point toward a registry function.
The second watchpoint is outbound mobility. If AFRINIC-issued resources face practical barriers to leaving the region that comparable resources elsewhere do not face, the market will price a regional discount. Evidence may appear in lower bids, more complex transaction structures, reluctance by buyers to accept AFRINIC-linked risk, wider spreads between sale and lease economics, or increased use of service contracts that avoid formal transfer. The discount may be visible before anyone names it.
The third watchpoint is inbound status. Africa's future IPv4 demand cannot plausibly be met only from AFRINIC's residual pool. If imported resources become subject to status conversion, future outbound limits or policy burdens that did not apply before entry, inbound supply will be discouraged. If AFRINIC makes imports safe, status-preserving and transferable, it will have chosen liquidity over symbolism.
The fourth watchpoint is leasing visibility. A constructive registry will recognise that operational delegation exists and will improve ways to record it. It will seek accurate contacts, routing-authorisation clarity, abuse-handling reachability and dispute metadata. A control-oriented registry will treat leasing mainly as suspicious commercialisation. The first posture improves the ledger. The second pushes the market into private paper and partial disclosure.
The fifth watchpoint is how fees and good standing are used. Ordinary billing discipline is legitimate. Using membership status to block unrelated transfers, impair reverse DNS, delay updates or create leverage in policy disputes is different. Because AFRINIC's fee structure is tiny relative to the market value of large IPv4 holdings, fee and membership tools can become high-leverage control points. Proportionality and cure periods matter.
The sixth watchpoint is whether abuse-contact policy stays thin. A registry can require reachable contacts and publish validation status. It should not evaluate the adequacy of every abuse response, become a universal complaint tribunal or use contact defects as a path toward resource impairment. Contactability is a ledger function. Abuse adjudication belongs to operators, customers, networks, courts and law enforcement.
The seventh watchpoint is institutional normalisation. AFRINIC's board, executive leadership, budgets, member records, bylaws, litigation posture and relationship with ICANN will all affect the risk premium attached to its policy choices. Governance recovery is not the same as economic reform, but weak governance makes discretionary conservation more expensive. The market will watch conduct more than statements.
The eighth watchpoint is whether poor-country rhetoric is converted into measurable pro-poor instruments. If the policy objective is African connectivity, the test is cheaper total access, more visible supply, lower transaction cost, better financing, improved local infrastructure and less dependence on institutional discretion. If the objective is simply to stop resources from leaving while maintaining a narrow pool and heavy review, the rhetoric will have become a cover for confinement.
The uncertainty should remain explicit. Courts may clarify parts of the Cloud Innovation dispute. AFRINIC may implement policies more narrowly than critics fear. Some transfer restrictions may be less damaging in practice than in theory. Fraud histories justify verification. IPv6 deployment could reduce some future dependence on IPv4, though dual-stack reality keeps IPv4 economically relevant. Some African operators may sincerely prefer regional retention if they fear being priced out by global buyers.
Even with those caveats, the main economic lesson is stable. Conservation is defensible when it protects uniqueness, accuracy and continuity. It becomes suspect when it masks choices about who may move scarce value, who may monetise underused capacity, who may import resources, who bears registry risk and who absorbs the discount from immobility. AFRINIC's challenge is not to say the word conservation more loudly. It is to prove that conservation no longer means control.
If the registry chooses liquidity, objectivity and narrow authority, conservation can regain its technical meaning. If it chooses regional retention, needs review and discretionary approval, the market will read the rhetoric in a simpler way. It will see a scarce asset behind a gate, and it will price the gate accordingly.