Summary
- What it says: AFRINIC is examined through consensus capture as a registry-governance and institutional-economics problem for the Africa region.
- Main topic: Network-resource evidence; Registry governance; Institutional legitimacy; Consensus capture
- Context: Governance / Research / Africa
AFRINIC's crisis is usually described through the visible drama: litigation, a board vacuum, receivership, contested elections, scarce IPv4 addresses and arguments about whether the registry can be kept operational. Those matters are real. They are also the noisy surface of a quieter economy. Beneath them sits a machine for producing consent. Mailing lists, the Policy Development Process, public meetings, working-group chairs, last calls, board ratification, member elections and the language of "community" do more than collect opinion. They convert scarce attention into institutional permission.
In a registry whose decisions can affect transfer mobility, business continuity and the market value of address resources, that permission has a price.
Consensus capture is not the same as ballot stuffing, bribery or a crude takeover. It usually looks more respectable. It can occur when ordinary procedure gives a small, persistent class of participants control over agenda, vocabulary, timing, evidentiary burden and closure while most resource holders remain operationally busy, poorly informed or rationally absent. The door may be open. The price of walking through it may still be too high. The archive then records activity, not necessarily consent.
AFRINIC makes the problem unusually visible because its policy process now operates where several pressures meet: permanent IPv4 scarcity, a contested registry model, litigation over enforcement and member rights, disputed election mechanics, and a transfer market that prices registry risk. AFRINIC's formal materials describe a bottom-up process for developing number-resource policy. It explains the machinery. It should not be allowed to become the frame through which every economic consequence is judged.
The harder question is whether the institutional record produced by that machinery is strong enough to justify binding consequences for absent holders, small operators, downstream customers and counterparties that never participated in the ritual. Heng Lu's public notes set out the test: a registry should protect uniqueness, accuracy, security metadata, transfer recording, fraud control, dispute isolation and continuity, but becomes dangerous when it moves from bookkeeper to gatekeeper and then toward something resembling a sovereign.
That distinction matters. If a registry acts alone, its discretion is visible and challengeable. If the same discretion appears after a proposal has moved through an open list, a public meeting, chair assessment, last call and board ratification, it can be presented as community will. The form is participatory. The effect can be distributive. The key question is not whether process happened. It is who paid the cost of participation, who controlled the terms of debate, who was absent, and who pays when the result changes resource-holder risk.
Consensus became valuable when IPv4 became capital
In the allocation era, consensus looked mostly like a rationing tool. A finite pool of unallocated addresses had to be distributed in a way that reduced duplication, preserved routing coherence and discouraged waste. Need assessment, conservation language, regional boundaries and public-resource vocabulary had an administrative logic. A policy room could argue about how to allocate something still held in a common pool. Applicants could decide whether to seek resources under the conditions on offer.
That world did not survive IPv4 exhaustion. AFRINIC's own materials record the region's movement into the later stage of its Soft Landing regime in January 2020. Once scarcity became permanent and address blocks acquired durable market value, policy ceased to be just a queueing rule for new applicants. It became part of the rule system around already deployed resources, routed networks, customer contracts, financing decisions and business models. A clause in a policy manual could change liquidity, exit options, compliance cost and the confidence with which a buyer, lender or customer treats a block.
The word "policy" therefore hides different asset classes. A consensus call about a contact format is not the same thing as a consensus call about transfer mobility. A discussion about future free-pool issuance can bind applicants more easily than it can bind existing holders who built networks under earlier expectations. A rule about abuse-contact reachability may improve accuracy; a rule that turns contact failure into broader enforcement leverage changes tail risk. The PDP vocabulary may be the same. The economics are not.
This is where consensus acquires market value. A block that can be transferred, financed, leased transparently and protected from discretionary review is worth more than a block whose future depends on a contested registry policy. Even if no resource is revoked, uncertainty creates a discount. Buyers price it. Lenders price it. Operators price it. Customers may pay it indirectly through higher prices, weaker continuity guarantees or less transparent contracting. The scarce resource is not only IPv4. It is accepted legitimacy over the rules under which IPv4 moves.
That is why the production of consensus deserves the same scrutiny as the production of policy. In a low-stakes technical environment, a thin record may be enough. In a scarcity environment, a thin record can transfer wealth, narrow exit, increase the value of board control and force private disputes into court. The mailing list and the policy meeting are no longer merely deliberative spaces. They are upstream instruments in a capital allocation system.
Every legally entitled Resource Member should receive registration and the applicable shareholder-equivalent protections; AFRINIC has no share capital. Cloud demands correction or removal of the purported directors’ entries through competent legal procedures. Its 2 November communiqué rejects the Receiver’s request for administrative removal copied to executive offices.
The institutional danger is not that AFRINIC has a policy process. It needs one. The danger is that a process designed for coordination can become a means of laundering discretionary control into a legitimate-looking output. The more valuable the underlying resource, the more valuable the appearance of consensus becomes. Capture then does not have to seize the registry from outside. It can work by controlling the conditions under which the registry is told that the community has spoken.
The open door is also a cost filter
AFRINIC's formal process is open in the familiar internet-governance sense. Policy proposals are posted, discussed on a public list, revised, taken to public meetings, assessed for rough consensus, put through last call and ratified if the process reaches the required stage. Anyone may participate. Archives are visible. Procedure can be described in neutral terms. It is not enough.
Openness answers an access question. It does not answer a cost question. A small ISP, a university network, a hosting company, a datacentre, a bank, a public body or a regional enterprise may have the formal right to subscribe to a list and object. That does not mean it has staff time to monitor every thread, confidence to argue in the accepted language, legal capacity to read downstream implications, institutional memory to understand old disputes, or incentive to engage before a proposal becomes concrete. Formal equality is cheap. Practical equality is expensive.
The labour rewarded by mailing-list influence is highly specialised. The effective participant reads long threads, remembers previous meeting minutes, cites manual sections, tracks versions, understands chair practice, returns during last call and knows which objections will survive. That is not the same skill set as running a network. Operators know outages, routing, procurement, abuse queues, peering, billing, customer dependency and service continuity. Governance regulars know timing, drafting, alliances, vocabulary and closure.
Expertise is not illegitimate. A registry policy process would be worse without people who can catch implementation defects and translate technical consequences into workable text. The problem is that procedural expertise can become a barrier to affected-party voice. A network operator may understand exactly how a transfer restriction, review power or compliance escalation affects its business and still fail to express the objection in a form the process treats as durable. A professional participant may understand less about that operator's exposure but more about how to win the archive.
This is capture by complexity. The more elaborate the policy environment becomes, the more valuable the people become who know how to navigate it. They may be registry staff, former insiders, consultants, lawyers, civil-society professionals, academics, policy advocates or unusually persistent members. The structural incentive is clear: complexity raises the return to incumbency. It lets process knowledge substitute for democratic breadth. It makes the people who know the maze more important because the maze remains difficult.
AFRINIC's wider crisis reinforces the filter. Independent reporting has described years of litigation, receivership, contested elections, proxy and power-of-attorney controversy, ICANN interventions, NRO statements, and arguments over bylaw reform and transfer rules. An ordinary member trying to understand a single policy proposal must often understand the legal and institutional weather around it. The result is predictable. Those who already live in the process remain. Those whose main work is operating networks arrive late or not at all.
The language of "community" then becomes slippery. It sounds broad, but the effective community in a high-cost process is usually the set of people who show up repeatedly. That set may include important expertise. It may not resemble the full population of resource holders and affected users. The archive is open, but filtered. It proves that participation was possible. It does not prove that affected parties were present, that objections were affordable or that silence meant acceptance.
Agenda control starts before the first objection
The strongest form of capture often begins before debate. The participant who defines the problem controls much of what follows. A proposal that frames resource mobility as leakage produces one kind of discussion. A proposal that frames it as liquidity produces another. A text that calls registry intervention continuity protection invites different objections from one that calls it discretionary control. A regional-retention proposal framed as fairness places critics on different terrain from one framed as an exit restriction.
In an open process, the agenda setter does not need to silence anyone. It is enough to provide the vocabulary within which everyone else must speak. Objectors are pushed into defensive work. They must prove that fairness may be unfair, that stewardship may be coercive, that regional protection may hurt regional operators, or that stability rhetoric can create instability. The initial framing gives proponents moral advantage and forces critics to spend scarce attention undoing the premise.
AFRINIC is vulnerable to this because policy vocabulary carries economic assumptions. "Public resource" implies common ownership or at least a collective claim. "Stewardship" implies guardian authority. "Regional" can turn geography into control. "Need" elevates administrative judgment over market demand. "Proper use" suggests a moral test beyond uniqueness and accuracy. "Community-developed policy" implies broad consent. These terms sound technical because they have been repeated for years. In scarcity conditions they become control words.
Heng Lu's "Policy Mirror" argument is useful for this reason. It treats policy language as evidence of institutional imagination. In that account, a narrow registry writes rules about holders, contacts, uniqueness, security assertions, transfer recording and dispute status. A sovereign-style registry writes rules about stewardship, regional entitlement, proper use, conservation, compliance, revocation and who may transact. Those words allocate burdens. Once a term is accepted, certain powers become easier to justify.
Timing is another agenda device. A proposal that sits in the archive for months can acquire the appearance of maturity even when participation was thin. A late objection may be treated as stale because the discussion has passed earlier stages. A resource holder who notices the economic effect only near implementation can be told that the correct moment to object has passed. The process rewards those present at inception and penalises those who become attentive when the cost becomes concrete.
The same dynamic appears in election mechanics. Board control does not write every policy, but it shapes ratification, staff posture, litigation strategy, bylaw reform, budget choices and the public narrative around disputes. If the board is produced through contested or opaque mechanics, later ratification is consolidation, not restored governance. The Register has reported on AFRINIC's boardless period, receivership, the June 2025 election (annulled by AFRINIC's Receiver on 26 June 2025), allegations involving voting authorisation and powers of attorney, and later efforts to restore board function. Cloud Innovation does not recognise the purported Board that followed (Cloud Innovation’s answers). They show that the institutional setting for policy ratification has been far from ordinary.
Agenda control is therefore not only a mailing-list matter. It links proposal language, PDP sequencing, chair judgment, election credibility, board authority and external recognition. A rule can win not because its text persuaded a broad membership, but because the institutional sequence made it difficult for opponents to find the right battlefield in time. That is why "the community discussed it" should be treated as the start of an inquiry, not the end of one.
Closure can become a procedural veto
Every governance system needs closure. Without it, a registry cannot operate. The difficulty is that closure devices can become procedural vetoes against economically exposed dissent. In AFRINIC's model, working-group chairs assess rough consensus, last call provides a final opportunity to object, appeals require procedural discipline, and the board ratifies. Each stage is defensible in isolation. Together they can create an expensive path for anyone trying to stop a consequential rule.
The proponent benefits from initiative. The objector's burden compounds. An objector must notice the proposal, understand the text, identify the economic harm, write the objection in the accepted idiom, stay engaged through revisions, follow or attend meetings, return during last call and possibly appeal. If the objection is late, it may be treated as re-litigation. If it is broad, it may be dismissed as ideological. If it is narrow, proponents may amend around it while preserving the main control point.
This is the procedural veto: not a veto held by one official, but a veto produced by friction. The process can say dissent had every opportunity. The dissenter experiences a sequence of costs that rises faster than the expected chance of success. Smaller operators, in particular, often stop paying before the process ends.
The appeal path illustrates the point. AFRINIC's policy manual describes a route for challenging chair action, including discussion with chairs and the working group and escalation through defined mechanisms. Such mechanisms are necessary. But an appeal that depends on timing, prior participation, procedural literacy and support from other participants will naturally favour insiders. A resource holder discovering exposure late may be more directly affected than a list regular, yet less able to trigger review.
Cloud Innovation went to court to protect its members’ rights and the networks using its addresses. The documented protections include the July 2021 restraint against freezing or reclaiming allocated resources and cutting off WHOIS access, and the September 2023 preservation orders and receivership directed towards an election under AFRINIC’s constitution. In 2022 SCJ 253, the Court rejected AFRINIC’s abuse-of-process objection to the application of 3 December 2021, recognising the new circumstances that required protection. Cloud’s operating objective is stable address use, accurate records and continuous registry services. Those instruments, not commentary, are the record of its legal response.
This creates a second-order cost for the whole system. If internal review is too insider-weighted, resource holders shift disputes to courts. If courts become the main check on registry discretion, operations slow and public narratives harden. The registry says it is being paralysed. Holders say they are defending continuity. The deeper defect is a policy process that did not create trusted, proportionate, affected-party review before conflict escalated.
A healthier system would make procedure a cost reducer, not a cost shifter. Serious economic objections should be easier to surface early, easier to test independently and harder to bury under consensus formulas. That would not give every resource holder a veto over technical housekeeping. It would prevent a small process class from imposing asset-governance consequences through a route designed for lower-stakes coordination.
Silence is not consent
The most dangerous input in consensus governance is silence. Silence is tempting because it simplifies decision-making. If a list is open, a proposal is public and few people object, it is easy to infer that the community can live with the result. That inference may be tolerable for minor operational changes. It is weak evidence for high-consequence resource policy.
Member apathy at AFRINIC should surprise no one. Many members are not policy institutions. They are network operators, access networks, universities, enterprises, public bodies and companies whose incentives are operational. They deal with outages, customers, suppliers, billing, abuse tickets, compliance and engineering. A policy thread is usually a low-salience risk until the policy enters implementation. By then the record may already show months of discussion and a procedural decision may already frame the matter as nearly settled.
Economists call this rational apathy. The cost of following every thread is immediate and certain. The chance that one member's intervention changes the outcome is uncertain and often low. The harm from a bad rule is distant, probabilistic and shared until enforcement becomes specific. Under those incentives, many rational members stay quiet. Their silence is then available to be repurposed as community acquiescence.
Heng Lu's notes repeatedly return to this point. He argues that many AFRINIC members do not realise how much voting or policy influence they formally possess, and that a disciplined minority can dominate when participation is low. The economic mechanism is ordinary. Low-turnout institutions are easy to steer. A small organised group can look like the community because the actual affected population is dispersed, busy and inattentive.
The silence problem is worse when a policy's economic effect is indirect. A rule may not say, in plain English, that it will reduce exit value or make financing harder. It may say resources are regional, transfers require approval, contacts must be validated, or compliance duties must be observed. The effect is cumulative. A holder may not see one clause as existential, but the stack of clauses can produce a risk premium: less mobility, more discretion, more uncertainty, more need for lawyers and less predictable continuity.
Silence can also reflect distrust. Some members may believe participation is useless because insiders have already decided the outcome. Others may fear public association with a faction. Some may lack confidence in English-language debate or in the etiquette of internet-governance argument. Some may depend commercially on parties inside the dispute and avoid visibility. Some may not understand that a policy labelled as technical can reach transfer value or resource recognition. None of these silences is consent.
A serious consensus system must therefore distinguish non-objection from informed acceptance. For policies that materially affect existing holders' transferability, portability, fees, revocation exposure, security services or recognised control, silence should count as a warning, not an asset. The process should ask why affected holders are absent, whether they received plain-language notice, whether the economic effect was explained, and whether they had a practical route to object. If the answer is no, the archive is a participation sample. It is not a mandate.
Elections are the back end of consensus capture
Mailing-list capture and election capture are often treated as separate problems. They are linked. Policy consensus supplies the rule text. Elections supply the board and institutional authority that turn text into action. A weak election can taint later ratification; a captured policy process can make control of the board more valuable. Each feeds the other.
AFRINIC's recent election history is central to that risk. Independent reporting has described a period without a functioning board, the appointment of an official receiver by the Supreme Court of Mauritius, attempts to arrange elections, a June 2025 vote that the Receiver annulled on 26 June 2025, with reporting pointing to concerns over voting authorisation and powers of attorney, and a later purported Board elected under continued scrutiny. Cloud Innovation’s position is that the Receiver annulled the vote over a single disputed proxy, banned the proxy voting that section 12.12 of AFRINIC’s constitution allows, and extended his own term without its knowledge; it does not recognise the resulting Board (Cloud Innovation’s answers). That is precisely why later claims of routine consensus deserve caution. A registry whose board legitimacy is disputed cannot treat ratification as administrative tidying.
Elections can be captured economically without falsifying every ballot. If members are apathetic, poorly informed or willing to delegate authority casually, organised actors can aggregate votes. Proxy systems are not inherently illegitimate; they help busy members participate. They become dangerous when one actor, faction, adviser or campaign can convert member inattention into concentrated control without strong evidence of informed authorisation. The same registry that must verify control over address records must also verify control over member voice.
The Register's reporting on the June 2025 controversy included questions about powers of attorney and alleged cases in which members disputed votes or authorisations recorded in their name. The institutional lesson is clear. A paper authorisation, a database entry or a credential can become a lever over scarce value. If authorisation is weak, the problem is not merely electoral. It is economic.
Board control matters because the board sits where policy ratification, management oversight, legal posture, budget authority, bylaw reform and public narrative meet. A board cannot create IPv4 addresses. It can influence how strictly policy is interpreted, how much discretion staff exercise, how aggressively litigation is fought, how transfer rules are presented, and whether the registry narrows or expands its own power. In a scarcity environment, that is a control premium.
Heng Lu's public note on AFRINIC lock-in argues that overwhelming electoral support in a fractured environment should be examined carefully rather than celebrated automatically, and that structural policy adopted while legitimacy remains unsettled can be read as consolidation. It raises the right governance question: when the source of board authority is contested, what kind of economic policy should the board avoid until its legitimacy is clearer?
The answer should be restraint. Routine services, publication, security, invoicing, staffing and continuity must continue. Irreversible or economically distributive policy - especially on transfer mobility, existing-holder rights, revocation exposure and bylaw changes affecting member power - needs a higher legitimacy threshold. Otherwise elections become the back end of consensus capture: first produce authority through low-participation or poorly verified mechanics; then use that authority to ratify rules said to reflect community will.
Vocabulary is a control surface
The most efficient capture mechanism is often a dictionary. If a registry or its allies can define the terms of legitimacy, they can shape the boundaries of acceptable objection. "Community" is the largest word in this dictionary. It can mean the full set of affected resource holders. It can mean meeting participants. It can mean mailing-list regulars. It can mean a region, a professional class, a technical tradition or a moral claimant invoked in public statements. The word expands and contracts according to need.
Heng Lu's public note on who speaks for a continent makes the point directly. He argues that an administrative service region is not a sovereign people and that a regional registry should not be confused with the continent it serves. The claim is analytically important. AFRINIC serves a region. That does not mean every policy participant speaks for Africa, every critic is anti-African, or every resource registered through AFRINIC is a regional trophy.
"Stewardship" performs a similar function. It sounds modest because it implies care. In practice it can smuggle in authority. A steward decides what proper use looks like. A steward may restrict transfer, judge hoarding, discourage speculation, police regional entitlement or elevate a moral claim over a commercial reliance interest. A bookkeeper records. A steward governs. The move from recordkeeping to stewardship is therefore not semantic decoration. It changes the kind of power being claimed.
"Stability" is another control word. Everyone wants registry stability. The question is what kind. A narrow stability frame means continuous publication, accurate records, reverse-DNS continuity, RPKI reliability, fraud control and dispute isolation. A thick stability frame can mean protecting the incumbent registry from challenge, discouraging litigation, preserving institutional discretion or treating exit as a threat. The same word can defend the ledger or defend the gatekeeper.
"Capture" itself is unstable. Incumbents may describe litigation, organised voting or reform campaigns as capture. Critics may describe insider dominance and procedural control as capture. The analytical test is not who uses the word most loudly. It is where decision rights sit, who bears the downside, what evidence supports authorisation, and whether affected principals can exit or obtain independent review.
This is why official narratives cannot be used as truth frames. The vocabulary of AFRINIC, ICANN, the NRO and other institutions is not neutral. A statement that registry stability is important does not prove that a given transfer restriction is proportionate. A statement that community process occurred does not prove informed holder consent. A statement that litigation threatens the registry does not prove the litigant lacks rights.
The vocabulary test is simple. Translate every institutional word into an operational question. Community becomes: which identified affected parties participated? Stewardship becomes: what power is being exercised over whose asset or service? Stability becomes: which concrete registry service would fail without this rule? Consensus becomes: who had notice, what did they understand, and what did silence mean? Capture becomes: who controls the agenda, record, vote, implementation and liability? Once translated, many slogans shrink. That is useful.
From archive to discretion
A mailing list is not a legislature. Yet in the RIR world it can become the evidentiary trail through which registry discretion is justified. A proposal is posted. Comments accumulate. Chairs assess consensus. Last call closes the record. A board ratifies. Later, when a member objects to implementation, the registry can point back to the process. The archive becomes a shield.
This conversion is the core economic act. It turns low-cost speech into high-consequence discretion. An email costs little. A policy manual clause can affect millions of dollars in address value and the continuity of services using those addresses. The conversion mechanism requires safeguards proportional to the stakes.
The AFRINIC transfer controversy illustrates why. Public notes by Heng Lu and reporting by The Register have described the significance of a ratified transfer framework, including the classification of AFRINIC-pool resources as regional for transfer purposes and resulting limits on outbound inter-RIR movement for many resources. AFRINIC has presented its policies in terms of regional governance, resource management and process. Critics describe the same framework as lock-in or capital control. Its economic importance is beyond dispute.
If a policy constrains transfer, it does not merely update a database. It changes bargaining power. A holder with global exit options can discipline a registry because it can move value, transact with a broader market or price alternatives. A holder locked into a narrower transfer environment depends more heavily on registry discretion and regional demand. That dependence changes the relationship between member and registry.
This is why the archive must not be treated as conclusive proof of authority. A discussion record may show that a proposal passed through the correct stages. It may not show that existing holders accepted a reduction in mobility. It may not show that downstream customers understood continuity risk. It may not show that small operators saw the financing implications. It may not show that the board ratifying the proposal enjoyed uncontested legitimacy.
KrebsOnSecurity's 2019 reporting on alleged historical address-record manipulation at AFRINIC provides a separate but related lesson. The allegations concerned a major address heist and the economic value of registry records. The reporting showed that registry databases are not harmless paperwork. Weak record controls around scarce IPv4 can create large economic harm. Policy records deserve similar seriousness because they shape the environment in which registry records are changed, contested or frozen.
The Internet Governance Project's work on the Cloud Innovation dispute also matters because it framed the conflict as a political-economy dispute rather than a mere compliance quarrel. That frame helps explain why litigation, transfer policy, scarcity, regional-use claims and member rights became inseparable. AFRINIC's policies operate in an economy. They can protect the ledger or expand registry power. Which they do depends less on official vocabulary than on actual consequences.
The most dangerous archive is one that becomes more authoritative than the economy it governs. A mailing list can record debate. It cannot, by itself, supply consent from absent principals. It can help a registry understand technical consequences. It should not become a mechanism for converting low-turnout process into discretionary authority over live networks.
The risk premium paid by resource holders
For a resource holder, consensus capture appears less as a headline than as a risk premium. The premium is paid in legal advice, delayed transfers, lower asset value, reduced collateral, contract uncertainty, reputational exposure and the need to monitor institutional politics that should have remained boring. A registry that should reduce transaction costs begins to add them.
The first premium is transfer risk. If policy restricts outbound movement, requires discretionary approval or allows broad review of business purpose, the holder's resource becomes less liquid. Liquidity is not a luxury. It is the mechanism by which capital finds higher-valued use. A block that can move globally commands a different price from a block whose movement depends on regional conditions and registry approval. Even if the holder does not plan to sell, lost optionality has value.
The second premium is enforcement risk. If compliance language can escalate from contact accuracy to contractual breach or service impairment, routine registry maintenance becomes a tail-risk event. Abuse-contact requirements may be justified as record accuracy. The danger appears when a thin contact rule becomes a thick enforcement tool. The holder then has to price not only whether it can receive notices, but whether a registry may use a process failure as leverage over resource recognition.
The third premium is governance risk. If board elections, member registers, proxies, powers of attorney or bylaw changes are contested, every later act by the board can be discounted or challenged. A buyer, lender, lessee or customer may ask whether a policy will survive litigation, whether a registry decision will be reversed, or whether a receiver, court, ICANN intervention or future board will alter the environment. The address may route normally while the administrative path around it becomes uncertain.
The fourth premium is participation risk. Resource holders must spend more time following policy threads, election notices, member credentials and litigation updates. This is a hidden tax. Large actors can hire lawyers, consultants and governance specialists. Smaller operators cannot. The result is regressive. The more complex and discretionary the registry environment becomes, the more advantage shifts to actors with capital and process fluency. Policies described as protecting less powerful networks may end up raising their cost of operation.
The fifth premium is narrative risk. If holders who object to policy are described as speculators, hoarders, outsiders, destabilising actors or threats to the community, ordinary commercial activity becomes reputationally charged. A leasing model, a transfer plan or a cross-border customer base can be recast as a moral defect. That discourages transparent transactions and pushes activity into informal structures. A registry that wants accurate records should fear this outcome. The database becomes less accurate when the official channel becomes dangerous.
Downstream users absorb part of the premium. Customers rarely know which RIR records the resources behind a hosting company, enterprise service or connectivity product. They notice when renumbering, routing reputation, reverse DNS, abuse handling or service continuity fails. If registry discretion threatens a holder, the operational consequences can travel to customers who never participated in the policy process and never authorised anyone to speak for them.
This is why resource-holder risk should not be dismissed as narrow private interest. Resource holders are the firms and institutions that turn numbers into working networks. Their balance-sheet risk becomes infrastructure risk when continuity is impaired. A consensus process that ignores their incentives is not more public-spirited. It is less realistic.
What an uncaptured process would require
An uncaptured policy process would not abolish mailing lists, public meetings or rough consensus. It would put them back in their proper place. They are instruments for discovering technical and operational judgment. They are not proof of consent by everyone affected. The larger the economic consequence, the more the process must add direct notice, impact analysis, affected-principal review and independent challenge.
The first reform is policy classification. AFRINIC should separate registry-mechanics policy from future allocation policy and existing-holder-impact policy. Data fields, publication formats, validation methods and security metadata can move through ordinary consensus if they do not impair existing interests. Future free-pool allocation can rely more heavily on ordinary process because applicants choose whether to request resources under published rules. Existing-holder-impact policy should face a higher threshold.
Existing-holder-impact policy includes rules affecting transferability, portability, leasing, commercial use, fees, service continuity, RPKI or reverse-DNS access, revocation exposure, dispute status and recognised control. Such proposals should trigger direct notice to affected holders, plain-language economic explanation, legal-authority analysis, anti-retroactivity analysis and independent review. Mandatory application to existing resources should require more than silence on a list.
The second reform is minority economics. Consensus reports should not hide serious unresolved objections behind formulas. They should identify the types of objections, the class of participants raising them and the likely operational or economic effect if the objection is correct. This does not give every objector a veto. It gives the board and members a record of what closure costs.
The third reform is authorisation hygiene. Elections and policy consultations should treat member voice with the same seriousness as resource control. Proxy rules should be uniform, limited, time-bound, specific, independently confirmed and visible to the member. A member should be able to confirm whether a vote or authorisation has been recorded in its name. Where the registry relies on member support for legitimacy, the evidence of member support must be auditable.
The fourth reform is independent review before crisis. A holder facing adverse action on transfer, portability, recognised control, reverse DNS, RPKI, publication or service continuity should have access to a review mechanism that preserves the last verified state while the dispute is examined. This would reduce the incentive to turn every conflict into broad litigation and prevent the registry from becoming judge, party and executioner.
The fifth reform is a narrow registry function. The mandatory layer should protect uniqueness, accuracy, contactability, fraud control, security assertions, transfer recording and continuity. It should not decide price, speculation, business model, customer geography, regional morality or capital destiny. Markets, contracts, courts, security communities and public authorities each have roles. The registry should not absorb them all under the word stewardship.
These changes would not weaken AFRINIC's useful authority. They would distinguish useful authority from discretionary power. A registry trusted because it is narrow, fast, auditable and neutral is stronger than a registry feared because it can turn low-participation consensus into broad control. The practical test is whether ordinary operators can understand the rule before it binds them. If they cannot, the process has failed no matter how many messages sit in the archive.
What the public record shows
The Register's coverage follows the institutional timeline: boardlessness, receivership, attempted elections, the Receiver's annulment of the June 2025 election, proxy and power-of-attorney controversy, later board formation, ICANN involvement, bylaw disputes and litigation. It shows that AFRINIC's policy environment sits inside a live legitimacy dispute rather than a settled administrative routine.
KrebsOnSecurity's reporting on the alleged African IP address heist shows that registry records around scarce IPv4 can be tied to very large economic stakes and allegations of internal weakness. The registry layer is not harmless clerical infrastructure once address blocks have market value.
The scope of a court measure must be read from its operative terms. An attachment or other interim financial measure is not proof that registry services stopped, and the sum claimed is not evidence of the registry’s actual bank balance. Equally, the interim order of 14 May 2026 (SC/COM/WRT/000338/2026), which AFRINIC’s Receiver obtained without notice against Cloud Innovation after a press release from LARUS, an independent company, restricts public statements. It is an interim order, not a final judgment; it does not decide the lawfulness of IPv4 leasing; nobody made the claim it prohibits; and a lawful business needs no court approval (what the order says). Cloud’s position is that each instrument must be applied to the issue it actually decides; unrelated proceedings cannot reopen its final membership inclusion direction.
Consensus capture does not require proof that every insider acted in bad faith. It requires only a process in which high-cost participation, agenda control, procedural closure, member silence and board uncertainty can produce rules whose consequences fall on people who were not meaningfully present.
Watchpoints for the next cycle
The first watchpoint is participation composition. Count not only the number of mailing-list messages but who writes them. Are small and medium operators present? Are large holders speaking directly? Are telecom operators, universities, datacentres, hosting companies and enterprises visible? Or is the discussion dominated by staff, former insiders, consultants, institutional allies, civil-society professionals and repeat governance participants? A thin but busy archive should not be mistaken for broad consent.
The second watchpoint is proposal origin and framing. High-consequence proposals should be read for the control point they create. Does the proposal define transfer as permission or recording? Does it treat geography as metadata or authority? Does it turn contact accuracy into enforcement? Does it convert regional rhetoric into asset restriction? Does it use stability to protect registry services or to protect institutional discretion?
The third watchpoint is chair reasoning. A serious consensus decision should explain why objections did or did not matter. It should separate preference from operational harm, legal uncertainty from inconvenience, and affected-holder risk from general ideological opposition. If a chair report simply declares that objections were addressed, it should be treated as weak evidence for high-consequence policy.
The fourth watchpoint is board restraint. If AFRINIC's board remains subject to legal or legitimacy dispute, it should avoid irreversible economic policy unless the authority is unmistakable and affected holders have had meaningful review. Restoring routine governance is not the same as acquiring a mandate to redesign resource mobility. A board that wants legitimacy should first make itself less valuable to capture.
The fifth watchpoint is implementation behaviour. The real policy is what staff do with the rule. Does AFRINIC use new policy narrowly to keep records accurate, or broadly to pressure business models? Does it publish clear reasons and appeal paths? Does it preserve RPKI, reverse DNS and last verified records during disputes? Does it isolate conflicts to the smallest affected resource, or does it let one dispute contaminate broader service?
The sixth watchpoint is rhetorical inflation. Whenever a participant says community, ask which people. Whenever someone says Africa, ask which members and which networks. Whenever someone says stability, ask which service. Whenever someone says stewardship, ask what power is being exercised. Whenever someone says capture, ask who controls the agenda, vote, record, implementation and liability. These translations are tedious. They are necessary.
The seventh watchpoint is market response. If AFRINIC-registered resources trade at a discount, if inbound transfers become unattractive, if leasing becomes less transparent, if lawyers become more central, or if operators seek alternative continuity mechanisms, the market is not being ideological. It is pricing governance. A registry can reject the language of assets, but it cannot prevent markets from pricing uncertainty.
The final watchpoint is whether silence is still being harvested. An institution that has learned from crisis will treat low participation as a legitimacy gap. An institution that has not will treat it as a free input. AFRINIC's future will depend less on whether it can recite bottom-up language than on whether it can prove informed, affected, auditable consent where existing holders carry the downside.
Consensus should lower the cost of coordination. When it raises the cost of exit, objection, participation and resource continuity, it has become something else. In AFRINIC's case, the risk is not that community process exists. The risk is that a small, procedurally fluent layer can convert community process into authority over people and networks that were never meaningfully present. That is consensus capture. Its economics are simple: the few pay the cost of participation; the many pay the price of the result.
Editorial correction. The 11 June 2025 Supreme Court order finally directed AFRINIC to enter Cloud Innovation as a Member under section 91. The 19 June category correction does not cancel that Member direction; Cloud remains an AFRINIC Resource Member and company-law Member. Administrative actors cannot retry or neutralise the decided inclusion; any stay, appeal or variation belongs to a competent court. See Cloud's rights statement and the separation-of-powers notice.
Editorial correction, 13 November 2026: the commentary framing has been replaced with the relevant court protections and Cloud Innovation’s rights position. This version also names the source of statements taken from AFRINIC’s Receiver and gives Cloud Innovation’s answer beside them. It also removes a description of how the piece was written, and analysis that excused AFRINIC’s management.






