AFRINIC is usually discussed as a litigation story, a scarcity story, or a recovery story after years without normal board governance. It is also a test of a quieter institutional question: when travel, visas, currency friction, language, time, employer permission, legal budget, proxy machinery and procedural fluency decide who can appear, how representative is the governance record that results?
The question is not whether AFRINIC's meetings, lists and elections are formally open. Much of the formal design points in that direction. The harder question is who can convert formal openness into usable influence. A small operator that cannot send staff, absorb a visa delay, pay for legal advice, speak confidently in English, follow years of list history or risk being named in a public fight is not present in the same sense as a repeat participant with money, lawyers and institutional memory. Both may have the same right to participate. They do not face the same price.
That price matters more because internet-number governance has become economically heavier than its technical vocabulary suggests. IPv4 scarcity, resource transfers, routing-security expectations, registry enforcement, member voting and litigation over address-control questions have turned procedural visibility into a commercial asset. In such a setting representation is not produced by invitation alone. It is produced by the distribution of participation costs.
The room is not the region
AFRINIC's official map is simple enough. The registry serves Africa and parts of the Indian Ocean. Its service region is divided into Northern, Western, Central, Eastern, Southern and Indian Ocean sub-regions for statistics and board elections. Its board is elected on a regional-representation basis, while directors are expected to act for the whole service region rather than merely for the seat from which they were elected. Its policy-development materials describe an open process in which anyone may participate, discussions are publicly archived and policy is developed through openness, transparency, fairness and consensus.
Those facts are important exhibits about the institution's design. They are not proof that the visible people in its processes are a good sample of the operator base affected by its decisions. A service region is not a parliament. A mailing list is not a census. A public policy meeting is not the African internet. A board seat labelled for a sub-region does not automatically equalise the cost of participation across that sub-region's members.
The distinction is practical, not semantic. AFRINIC allocates and registers IPv4 addresses, IPv6 addresses and autonomous system numbers. Its records support WHOIS, RDAP, reverse DNS, routing-security functions, abuse contacts, transfer confidence, member standing and operational continuity. In the IPv4 scarcity era those records are not merely entries in a technical database. They are the recognised control layer around scarce address capacity used by networks to serve customers, obtain upstream connectivity, meet security obligations and maintain commercial confidence.
When the registry's decisions were mostly about allocating a low-priced coordination resource from an expanding pool, the mismatch between formal openness and actual representation was easier to tolerate. If a small operator missed a meeting, the next allocation or the next policy season might still offer room for adjustment. Scarcity changed that bargain. A policy about transfers, usage, need, eligibility, revocation, proof of control or dispute handling can now move value among holders, buyers, lessors, users and future entrants. The cost of not being represented is no longer merely civic. It can become financial.
The relevant boundary is therefore representation itself: who can afford to be visible in governance. The basic problem is not simply how issues are queued, how meetings are chaired, or how a single procedural step is interpreted. It is that the names appearing in minutes, election campaigns, list threads and court exhibits are the names of people and organisations able to pay the participation price at the relevant moment. The affected base is wider.
That base includes national and regional internet service providers, small access networks, data centres, wireless operators, universities, Internet exchange points, government networks, hosting companies, banks, software platforms, civil-society networks, enterprise networks, legacy address holders, IPv6 advocates, IPv4-dependent businesses and customers who never speak in registry proceedings at all. Some have lawyers and dedicated governance staff. Some have one engineer who handles routing, billing, outages, procurement and regulator correspondence. Treating those groups as equally present because the door is formally open is a category error.
AFRINIC is the hardest regional case because the representation question became visible under stress. The Cloud Innovation dispute, the long period without a normally functioning board, receivership in Mauritius, the annulled June 2025 election, allegations around powers of attorney, a later board restoration, continued litigation, ICANN intervention, IPv4 scarcity and competing reform campaigns all exposed the same hidden variable: the ability to participate is unevenly distributed. In a crisis, the people who can keep showing up become the institution's visible public. That does not make them illegitimate. It means their visibility should not be confused with full representation.
Openness has a price
The most reassuring phrase in registry governance is "anyone may participate". AFRINIC's policy materials use that idea explicitly. Policies are developed in an open forum with no qualification for participation; the Policy Development Working Group is open through the internet or in person; anyone can submit a proposal; proposals move through list discussion, public policy meeting, last call and board ratification. In a formal procedural sense, this is openness.
In institutional economics, however, openness is not the same as access. A market may allow anyone to trade, but the trader still needs capital, information, settlement capability and risk tolerance. A court may allow any claimant to sue, but the claimant still needs filing fees, counsel, time and evidence. A public meeting may allow anyone to speak, but the speaker still needs to know that the meeting exists, understand the rule at stake, trust the forum, read enough history, obtain permission to speak, overcome language costs, absorb the time burden and accept the social risk of a public position.
The real test is the price of converting a formal right into usable influence. In AFRINIC's case that price has many components. A participant must follow technical and policy vocabulary. They must know which list matters and which discussion is merely noise. They must understand how a draft changes from one version to the next. They must identify when a meeting is decisive rather than informational. They must distinguish policy text from staff implementation practice. They must understand the relationship between community discussion, board ratification, corporate bylaws and Mauritian legal reality. During the crisis years, they also had to follow court orders, receiver communications, ICANN letters, competing public campaigns and claims about election rules.
That is not a low-cost environment. It favours repeat participants who have already paid the learning cost. It favours large operators with governance staff. It favours consultants who can turn accumulated procedural knowledge into a service. It favours advocacy organisations that monitor multiple venues. It favours lawyers and address-market specialists whose commercial stakes justify sustained attention. It disfavours the ordinary operator who discovers a rule only when a resource request, transfer, compliance review or vote suddenly matters.
The result is not necessarily conspiracy. It is selection. Formal openness lets many actors enter, while participation costs decide which actors keep entering. This is why the economics of representation cannot be settled by pointing to public archives, open meetings or remote participation. Those features lower some barriers. They do not remove the fixed cost of understanding what is consequential.
An engineer at a small ISP may be free to join a discussion in theory, yet find the right unusable in practice if the thread requires reading years of past debate, interpreting legalistic transfer language and exposing the employer's business concerns in a public forum. A university network may have a genuine view on documentation burdens but no one authorised to state it. A public-sector participant may be able to attend a meeting but not to speak without ministry clearance. Many rational actors will remain quiet or delegate the problem to someone else.
Openness should therefore be treated as an input, not a conclusion. AFRINIC's official process is more open than a private club. But the economic question is narrower and more demanding: after all the costs are paid, who remains visible, and what interests do they overrepresent?
Travel turns money into voice
AFRINIC holds open public policy meetings in different locations across its service region. The rotation is meant to distribute access and avoid permanent capture by one city. In some respects it does. A meeting in Nairobi is easier for a Kenyan operator than one in Tunis. A meeting in Mauritius is easier for an Indian Ocean participant than one in Dakar. Moving the meeting recognises that Africa is not one logistical unit.
But rotation does not abolish travel cost. It redistributes it. The same meeting that lowers the price for one group raises it for another. Air routes within Africa are often expensive, indirect and unpredictable. A journey between two African cities may require a connection through a third region or a long layover. Visa requirements differ by passport, employer, destination and notice period. Hotel costs, airport transfers, daily subsistence and time away from operations can exceed the visible registration price. A small access provider may need the same engineer who would attend the meeting to remain at home because the network has no spare team.
The official record may show that a meeting was open and that participants came from several countries. It will not necessarily show the operators who declined because the fare was too high, the visa process too uncertain, the employer too cautious, the meeting notice too late or the opportunity cost too large. Absence is rarely itemised. The record sees the room, not the shadow price of entering it.
Travel also creates a social advantage. The person who attends in person hears corridor discussions, meal conversations, side explanations and informal warnings. They see who is aligned with whom. They learn whether a proposed objection will be welcomed, ignored or socially punished. They can ask a staff member to clarify implementation details after a session. They can build credibility by being physically present. A remote participant can hear the microphone if the technology works. They do not necessarily share the room's social information.
In a registry crisis, that information becomes valuable. When board seats, resource transfers, legal strategy and institutional survival are all under discussion, informal knowledge helps participants decide whether to speak, whom to trust and where the real decision is happening. Those who can travel repeatedly become not merely better informed but better connected. They become more likely to be quoted, approached, endorsed or asked to serve. Their views start to look like the community's centre of gravity because they are closer to the venues where gravity is measured.
The June 2025 election illustrated the problem in sharper form. The election combined online voting and an in-person voting day. The in-person session became the focus of allegations around powers of attorney and voter documentation. South Africa's Internet Service Providers' Association told The Register that some authorised representatives encountered claims that another person had voted or could vote on their behalf. ICANN raised questions about election integrity and proxy rules. The receiver annulled the vote after concerns about voter documentation.
Those allegations must be handled conservatively. A public analysis cannot decide which power of attorney was valid, which was invalid, or which claim reflected confusion rather than misconduct. But the episode shows the weakness created when participation costs are high enough that representation is often mediated. If many resource holders cannot, will not or do not know how to appear directly, the value of proxies, credentials, endorsements and vote aggregation rises. A voting system must then distinguish genuine representation from brokerage. That is hard in any membership body. It is harder where members are dispersed across a large region, where the stakes include scarce IPv4 records, and where litigation has made every procedural error commercially explosive.
Travel cost therefore has two effects. It excludes some actors directly and creates a market for intermediaries who can appear on their behalf. The first effect makes the visible room unrepresentative. The second can make representation itself contested. AFRINIC's crisis did not invent those dynamics. It exposed them.
Visa friction, currency friction and the employer veto
Participation costs are not only monetary. They are administrative. A network operator may have the technical skill to contribute and the commercial exposure to care, yet still be stopped by passport restrictions, foreign-exchange controls, procurement procedures, security approvals or an employer that sees internet-governance travel as optional.
Visa friction matters because it is unequal by nationality. A participant from one country may book late and fly. Another may need an invitation letter, bank statements, hotel proof, employer confirmation, a consular appointment and a wait that does not fit the meeting schedule. A third may be able to travel only through a transit country with its own requirements. These frictions do not appear in consensus minutes. They shape who can attend before consensus is ever tested.
Currency and payment friction matter too. Flights, hotels and conferences are often priced in currencies that smaller African firms may not access quickly or predictably. Expense reimbursement may take weeks. A small company may not want to carry the cash-flow risk of sending a staff member to a meeting whose business value is hard to explain to accounting. A public institution may require procurement approval that closes after the relevant policy deadline. A university may be unable to fund travel outside an academic calendar. A civil-society participant may depend on fellowship support, which affects independence and availability.
Then there is the employer veto. Participation in AFRINIC governance is rarely someone's only job. The people most affected by registry policy are often employed to keep networks running, not to spend days on list debates. Their employer may allow travel for technical training, peering, vendor negotiation or customer acquisition, but not for an open-ended policy meeting. Even if permission is granted, the employee may be told to listen rather than speak. Public comments can bind or embarrass an organisation. They can reveal commercial dependency. They can attract legal attention. In a region where the AFRINIC dispute has involved litigation, public accusations and political claims, caution is rational.
This is one reason formal community processes overcount the views of people whose organisations let them speak freely. Consultants, independent advocates, association officials and senior executives may speak more often because public positioning is part of their role. Engineers inside smaller operators may have better operational knowledge but less permission to use it in public. A policy record that treats speech as a voluntary expression of preference will exaggerate the representativeness of the speakers.
The employer veto also creates a hierarchy between technical and legal participation. An engineer may not be authorised to express a view on legal liability, transferability or ownership-adjacent rights. A lawyer may not understand routing operations well enough to discuss implementation. A chief executive may care about business risk but lack the detail to comment on policy text. Larger firms can send all three or coordinate them internally. Smaller firms cannot. The result is not just fewer voices, but narrower voices.
AFRINIC's litigation environment magnifies the problem. When Cloud Innovation, AFRINIC, ICANN, NRS, LARUS, industry associations and other actors make public claims about authority, misuse, paralysis, liquidation, continuity or capture, ordinary members must decide whether speaking will pull them into a conflict they cannot afford. A small operator may believe a transfer restriction is harmful, a proxy arrangement suspicious, or a registry enforcement claim too broad, yet decline to comment because the reputational and legal costs are unclear. That absence should not be read as indifference.
Good representation design begins by recognising these hidden vetoes. If the only way to be counted is to travel, speak publicly, join a list, master the archive and accept legal visibility, the process selects for organisations that can absorb those costs. It does not necessarily select for those most affected.
Language and dialect are governance instruments
AFRINIC's language position is more complicated than a simple complaint about English. Its website offers several language options and its service region contains English-speaking, French-speaking, Arabic-speaking, Portuguese-speaking and many other linguistic communities. Some public materials are translated on a best-effort basis. At the same time, the English version is treated as official on the site, and the Consolidated Policy Manual states that communication with AFRINIC will be in English.
This is understandable administratively. A registry needs an authoritative text. Staff cannot operate every contractual, technical and legal interaction in every language of the service region. The internet's technical vocabulary is heavily Anglophone. Many African operators already work in English for routing, vendor, standards and upstream purposes. A single operating language reduces ambiguity.
But it also prices participation. English fluency is not the same as policy fluency. A participant may configure BGP in English, read vendor documentation in English and still lack confidence to challenge a draft policy in a public, archived debate. A lawyer may read English but not know the technical dialect. A French-speaking regulator may understand the governance issue but hesitate to intervene in a fast-moving list discussion. An Arabic-speaking or Portuguese-speaking operator may write a short operational question but not a careful objection about liability, transfer restrictions or consensus thresholds.
Language cost appears in timing as well as grammar. The first version of a proposal shapes the debate. The first objections create the frame. The first summary tells later readers which issues matter. If a participant waits for translation, internal explanation or help from a bilingual colleague, the social meaning of the proposal may already have settled. By the time they speak, they may appear late, repetitive or marginal. The same argument would have had more force if made by a fluent repeat participant on day one.
Every governance system also develops a dialect. In AFRINIC's case the dialect includes "rough consensus", "last call", "resource member", "registered member", "soft landing", "need", "utilisation", "reverse delegation", "LIR", "PI", "PA", "sub-allocation", "assignment", "good standing", "inter-RIR transfer", "abuse contact", "PDWG", "NomCom", "proxy", "power of attorney", "receiver" and "ICP-2". None of these terms is inherently improper. Specialist vocabulary is necessary for specialist institutions.
The representation problem begins when fluency in the dialect becomes a proxy for legitimacy. A participant who uses the expected vocabulary sounds serious. A participant who describes the same issue in ordinary business language may sound imprecise. "This blocks our ability to serve customers outside our first country" may be treated as weaker than a carefully framed claim about regional-use interpretation, transfer policy, RSA obligations and proportionality. Yet the business sentence may contain the more important fact.
Policy dialect can hide distributional choices. "Demonstrated need" sounds technical, but it can favour organisations whose future demand can be documented in the form expected by hostmasters. "Regional use" sounds developmental, but it can restrict networks whose customers, infrastructure or commercial relationships cross borders. "Resource stewardship" sounds public-spirited, but it can expand discretion if not tied to objective registry functions. "Community consensus" sounds democratic, but it may summarise the views of a small group able to participate.
The Cloud Innovation dispute shows why dialect is not harmless. The case turned partly on interpretations of usage, need, region, membership, contractual obligation and registry remedy. Internet Governance Project's 2021 analysis argued that AFRINIC overreacted to earlier problems and treated out-of-region use and changing use as grounds for an aggressive remedy. AFRINIC and its supporters framed the matter as policy enforcement and protection of regional resources. Lu Heng and related organisations framed it as a structural problem of registry power over scarce economic resources. Each side used a different dialect to describe the same control surface.
For ordinary members, the dialect war raises participation cost. To intervene intelligently, they must know not only their own operational exposure but the language in which the institution will recognise that exposure. If they say "we need flexibility", the answer may be "policy requires justified need". If they say "we own this block", the answer may be "resources are not property". If they say "this is a market", the answer may be "the registry is a custodian". If they say "this is Africa's resource", the answer may be "routing and customer demand are global". Each phrase carries institutional baggage.
A more representative system would translate dialect back into consequences. Every consequential proposal should answer plain questions: who would have to do something new, who would lose an option, who would gain a veto, what records would change, what evidence would be required, what timelines would apply, what happens to existing holders, and who can appeal. Plain-language summaries, French and Portuguese executive summaries before decisive stages, multilingual explanation of practical effects and a public record of when summaries became available would not make AFRINIC a multilingual legislature. They would make language a measured governance cost rather than an unspoken filter.
The core point is conservative. A registry that hears weakly phrased objections before they become litigation is stronger than one that waits for fluent repeat players to define the record.
Time is an operating cost
The internet-governance class often treats time as civic contribution. Operators experience it as cost. Every hour spent reading a policy thread is an hour not spent on outages, peering, customer provisioning, billing disputes, regulator forms, procurement, cable cuts, abuse complaints or staff training. In a large organisation those hours can be assigned to a policy specialist. In a small one they come from the same people who keep the network alive.
AFRINIC's policy process contains formal timing rules: draft policies are posted before meetings, agendas are announced, late changes are restricted, last call runs for a minimum period, minutes are published and appeals have deadlines. These rules create order. They also create time markets. The participant who can read immediately has more influence than the one who discovers the issue at last call. The participant who can compare versions has more influence than the one who reads only the current draft. The participant who can attend both the list and the meeting has more influence than the one who can do only one.
Time-zone costs are less dramatic than travel costs but more persistent. A remote session scheduled conveniently for the meeting room may fall during network maintenance, school runs, prayer time, local peak traffic or the working day of a participant who cannot multitask. A mailing-list thread may move while a small operator is handling a weekend outage. A deadline may coincide with a national holiday in part of the service region. None of these frictions is individually disqualifying. Together they select for people whose jobs allow them to treat policy attention as flexible.
Attention also depreciates. A participant who reads a proposal in April may need to reread it after revisions in June, after comments in July, after legal analysis in August and after implementation notes later. The cost is not just initial attention. It is repeated reacquisition. Repeat players carry context across cycles. They remember why a phrase was inserted, which compromise made a proposal acceptable, who objected before, which staff concern mattered and how a similar issue ended last time. Occasional participants face the full cost each time.
This is why institutional memory is a form of capital. It reduces the marginal cost of future participation. A person who followed AFRINIC through resource-use disputes, the Cloud Innovation litigation, receivership, board-election failure, ICANN interventions, transfer-policy debate and operational recovery can read a new controversy quickly. A resource member whose attention begins with a specific notice must reconstruct years of context. The first person can speak with confidence. The second may remain quiet because they cannot tell whether their concern has already been answered.
The cost of time also changes whose interests are visible. Actors with concentrated stakes monitor more closely than actors with diffuse stakes. A company holding large IPv4 resources will rationally track every policy and legal move. A broker whose business depends on transfer liquidity will do the same. A regional advocacy organisation can assign staff to the issue. A small access provider with limited staff may be affected by the same governance rules but cannot justify constant monitoring. The process then overrepresents concentrated stakes even when the formal rule is open participation.
This does not make concentrated participants wrong. Often they identify problems early because they have the incentive to pay attention. Cloud Innovation, NRS and LARUS have repeatedly forced attention onto registry power, liability, asset use and continuity. AFRINIC and its supporters have forced attention onto record integrity, regional stewardship, organisational survival and policy compliance. Both clusters have interests. Both also bring information. The point is not to disqualify them. It is to avoid mistaking their capacity to sustain attention for evidence that the wider affected base has been represented.
Time is the least visible participation cost because it leaves no invoice. It should be treated as one of the most important.
Public speech is not equally safe
Open governance assumes that participants can speak. Institutional economics asks whether speaking is safe enough to be rational. In AFRINIC's environment, that question cannot be brushed aside.
Public participation creates a durable record. Mailing-list comments are archived. Meeting statements may be minuted or recorded. Candidate endorsements circulate. Court exhibits and media reports can quote public positions. In calm times this publicity supports accountability. In contested times it creates risk. A member that criticises a registry policy may worry about future dealings with the registry. A member that criticises Cloud Innovation or NRS may worry about legal attention or public rebuttal. A member that criticises ICANN may worry about appearing hostile to the global coordination system. A member that supports a controversial candidate may worry about reputational association.
These fears are not symmetrical. A large company can absorb public disagreement. A small operator may depend on a few upstreams, local licences, government relationships and customer trust. A public university may not want its engineer to express a position that could be interpreted as institutional policy. A bank may not want to disclose its dependence on particular address records. A government network may avoid public statements that cross the line between technical administration and state policy. A civil-society participant may fear being treated as aligned with one faction after making a narrow procedural point.
Litigation raises the speech price further. The AFRINIC controversy has generated years of lawsuits, public accusations and allegations about capture, paralysis, fraud, proxy campaigns and institutional misconduct. Even where a particular member is not a party, the atmosphere changes incentives. A cautious organisation will ask why it should speak publicly if the same concern can be handled privately, through an association, or not at all.
The process-purist answer is that governance needs public records. That is correct but incomplete. A public record that captures only those able to tolerate exposure is not fully representative. It may overrepresent activists, repeat players, industry associations, litigants, consultants and institutional officials. It may underrepresent the quietly exposed. This is especially likely in a membership body where many participants have commercial relationships with one another.
Public-speaking confidence is another undercounted cost. Some people can take a microphone and disagree with a chair, a lawyer, a chief executive or a well-known community figure. Others cannot. Confidence is shaped by language, seniority, gender, age, institutional culture, previous treatment and familiarity with the room. A junior engineer may see the implementation problem earlier than a manager but have no authority to say so. A woman in a male-dominated technical room may face a different social price for persistence. A participant from a smaller market may be reluctant to challenge a speaker from a larger association or better-known operator.
These costs matter because AFRINIC's governance often relies on interpreting the weight of objections. A single objection from a confident repeat participant may be treated as substantive. Several weaker expressions from less fluent participants may be treated as questions. A process can be procedurally open and still hear confidence as if it were representativeness.
There is no perfect fix. Secret policymaking would be worse. But representation can be improved by adding structured low-risk channels: confidential impact submissions summarised without names, sectoral listening sessions, post-meeting surveys tied to specific proposals, published issue summaries that include anonymised concerns and explicit statements about how non-public feedback was used. Such channels should not replace public debate. They should correct its selection bias.
In infrastructure governance, the aim is not to make everyone loud. It is to ensure that quiet exposure is not invisible.
Representation becomes a brokered commodity
When direct participation is costly, representation becomes a commodity. Someone else offers to monitor, interpret, speak, endorse, vote, advise, aggregate, mobilise or litigate. The intermediary may be an industry association, law firm, consultant, broker, advocacy organisation, government-backed body, regional initiative, technical community group or informal network of repeat participants. Some intermediation is valuable. It lets small members pool attention and expertise. It also creates the risk that the intermediary's interests become hard to distinguish from those of the represented.
AFRINIC's 2025 election record made this unusually concrete. The receiver's April 2025 election plan appointed senior British lawyers to oversee nominations after concerns about potential interference. South Africa's ISPA had warned members earlier to protect AFRINIC credentials. The June election was later suspended and annulled amid concerns over powers of attorney and voter documentation. The Register reported claims that one party asserted authority to represent a large share of resource holders, along with allegations that some documents were fraudulent or could not be produced. ICANN asked questions about differences between online proxy rules and in-person power-of-attorney arrangements.
Again, the validity of specific documents is a matter of evidence. The broader economic lesson does not depend on deciding every allegation. If members cannot cheaply and confidently participate themselves, the right to participate becomes something others try to collect. Credentials, proxies, endorsements and powers of attorney acquire strategic value. The visible vote may then reflect not only member preference, but the ability of organised actors to reduce the participation cost for some members while increasing those members' dependency on the organiser.
This is not unique to AFRINIC. Trade associations, unions, shareholder meetings and political parties all face similar problems. But AFRINIC is a high-stakes case because the institution controls the recognised record around scarce resources. Board composition can affect budgets, litigation posture, bylaw reform, staff appointments, transfer policy, enforcement priorities and relations with ICANN and the other registries. The value of representation rises with the value of the registry's decisions.
Intermediation can also be epistemic. Many members cannot assess a policy or legal claim independently, so they rely on a trusted organisation to tell them what it means. Smart Africa, South African network-operator groups, ISPA, NRS, LARUS, ICANN, AFRINIC staff, lawyers and media outlets have all helped different audiences interpret the crisis. Each carries a perspective. Some are directly interested; some are institutionally invested; some are external observers with their own assumptions. Their interpretations reduce information costs but can also steer attention toward a preferred problem.
The later 2025 board election, after the annulled June attempt, further illustrated organised visibility. AFRINIC announced eight directors, and The Register reported that seven of the eight elected directors had Smart Africa's endorsement. That does not make the election illegitimate. Coordinated endorsements are a normal feature of contested governance. But it shows how organised actors can become decisive when many individual members face high participation costs. A slate is a participation shortcut. It tells members who lack time to study every candidate: vote this way if you trust us.
Slates can help when they discipline a fragmented field. They can be dangerous when they turn representation into packaged authority. The test is not whether a slate exists, but whether members can see who is behind it, what interests it reflects, what trade-offs it hides and how dissenting candidates can be evaluated without paying excessive research costs.
AFRINIC's future representation problem will not be solved by banning intermediation. That would hurt small members most. It will be solved by making intermediation auditable: clear proxy rules, direct member confirmation, vote receipts, public endorsement disclosures, conflict declarations, accessible candidate comparisons, spending transparency where relevant and easy challenge mechanisms before results become institutionally explosive.
Representation will always be partly brokered in a dispersed technical community. The question is whether the brokerage is visible enough to be trusted.
Repeat players do not need capture to dominate
The phrase "capture" is often too dramatic for what happens in registry governance. A process can be dominated without being captured. It is enough for repeat players to face lower participation costs than everyone else.
Repeat players know the lists, the chairs, the staff, the calendar, the legal background, the old compromises, the likely objections, the private anxieties and the difference between a symbolic debate and a decisive one. They can recycle arguments across proposals. They can recognise a small wording change that matters. They can draft amendments quickly. They can mobilise supporters. They can challenge minutes. They can file appeals. They can brief journalists. They can appear at successive meetings. They can treat the process as a portfolio of related interventions rather than isolated events.
That advantage compounds. An occasional participant may spend the first intervention merely proving that they understand the process. A repeat participant starts beyond that point. They may not be smarter or more legitimate. They have amortised the fixed cost of participation.
AFRINIC's crisis created several repeat-player classes. Registry insiders and long-term community figures carried institutional memory from the pre-crisis era. Resource-market actors and their advisers carried commercial urgency and legal strategy. ICANN and the NRO carried global-system continuity concerns. African internet-governance organisations carried regional-development and institutional-survival frames. Journalists and outside analysts carried narrative continuity. Lawyers carried court memory. These groups disagreed, but they all had more capacity to remain present than ordinary members.
The repeat-player advantage is not necessarily bad. Infrastructure institutions need memory. Newcomers can be naive about fraud, routing security, abuse handling, legal continuity and the economics of scarce resources. A process composed only of first-time participants would be unstable. But a process that lets repeat-player visibility stand in for representation will gradually narrow its reality.
One symptom is the recycling of moral categories. AFRINIC's debate often presents actors as defenders of Africa, profiteers, institutional loyalists, market radicals, community guardians, litigants, attackers or saviours. Those categories are easy for repeat players because they compress years of conflict into labels. They are costly for ordinary members because they make narrow operational positions look like factional alignment. A small operator that wants clearer transfer rules may not want to be labelled as supporting export of African resources. A member that wants continuity of AFRINIC services may not want to endorse every official enforcement claim. A participant that worries about proxy fraud may not want to be pulled into a broader campaign against one side.
Another symptom is issue bundling. A participant may care about lower documentation burdens but not support deregulated leasing. They may support stronger fraud controls but oppose broad revocation discretion. They may want ICANN to preserve continuity but not centralise authority. They may favour regional development but not address lock-in. Repeat-player campaigns often bundle these questions because bundles mobilise. Representation requires unbundling them so occasional participants can express partial agreement without joining a faction.
The best antidote is structured disagreement. Instead of asking who is for or against AFRINIC, a representative process should ask narrower questions. Does this proposal increase documentation cost for small operators? Does it change transferability for existing holders? Does it create a new staff discretion? Does it protect against forged authority? Does it impose regional-use assumptions on customers? Does it give members enough notice? Does it rely on English-only materials? Does it require travel or legal advice to understand? Such questions reduce the advantage of those who profit from broad narratives.
Repeat players will always matter. The institutional design challenge is to prevent their lower costs from becoming an invisible franchise.
Legal budget is political capacity
AFRINIC's recent history cannot be separated from law. The Cloud Innovation dispute moved from a resource-use and contractual argument into years of litigation, injunctions, bank-account freezes, board paralysis, receivership, election disputes, attempted liquidation and continuing claims. Courts in Mauritius became a central venue for registry governance. ICANN and the other registries were pulled into continuity questions. Public arguments about whether addresses are property, contract rights, public resources or economic assets acquired legal and commercial significance.
In such an environment legal budget becomes political capacity. The actor that can file, defend, appeal, seek injunctions, review bylaws, challenge nominations, prepare powers of attorney, advise voters, interpret court orders and brief the public has more influence than the actor that can only complain after the fact. This is not a moral judgment. It is an institutional fact.
Legalisation raises the cost of representation for ordinary members. A member considering whether to vote, appoint a proxy, support a candidate or object to a policy may need advice on corporate authority, resource-member status, registered-member status, bylaw procedure, insolvency supervision, proxy form, power-of-attorney validity and exposure to challenge. The cost of being wrong can be high. A disputed vote may be invalidated. A public claim may be rebutted. A document may be challenged. A member may be named in correspondence. Many will rationally avoid participation unless someone else pays the legal interpretation cost.
This can produce two opposite distortions. The first is withdrawal: members stay away, leaving the field to litigants and institutions. The second is delegation: members assign their voice to an intermediary that supplies legal confidence. Both reduce direct representation. Both make the visible record a product of legal capacity as well as member preference.
Legal budget also shapes narrative. AFRINIC has argued that litigation and procedural roadblocks have obstructed its ability to restore normal operations and serve members. Lu Heng and related organisations have argued that litigation exposes a structural problem: registries exercise high-consequence power over economically critical resources while bearing limited liability. Internet Governance Project has criticised both AFRINIC's aggressive resource posture and Cloud Innovation's excessive legal tactics. ICANN has framed intervention around continuity and global coordination. These arguments are not merely public relations. They are legal-economic theories about who should bear risk.
The ordinary member may agree with parts of several theories. They may think AFRINIC should not be paralysed by lawsuits, that Cloud Innovation deserved due process before any resource withdrawal, that forged voting authority would be intolerable, that ICANN should preserve continuity but not override local law casually, that transfer rules should be objective and that regional development cannot be reduced to institutional survival. Expressing that mixed position in a legalised environment is difficult. It is easier to avoid the fight or follow a trusted slate.
For representation, the lesson is uncomfortable. A membership body's legitimacy cannot rest solely on formal voting rights when the cost of understanding and exercising those rights includes legal advice. The institution must make the legally relevant parts of participation cheap and standardised: member-authority verification, proxy limits, plain bylaw explanations, candidate eligibility criteria, voting receipts, appeal windows, dispute categories and clear separation between court-ordered constraints and ordinary governance choice.
Reducing legal cost does not weaken the rule of law. It prevents law from becoming a private tollbooth into governance.
The visible participant is not the affected operator
The main representational fallacy is to equate visible participants with affected operators. In AFRINIC's case the affected set is broad and layered. It includes the resource member, the downstream ISP, the end-user network, the customer whose service depends on address continuity, the upstream that filters routes, the abuse desk that relies on contact data, the buyer or lessor evaluating registry risk, the bank financing infrastructure, the government using public networks and the small business that just needs reachable service. Most of these actors will never speak in AFRINIC proceedings.
This is not unusual. Infrastructure governance often acts through intermediaries. Electricity regulators hear utilities more than households. Port authorities hear shipping lines more than consumers. Securities registrars hear brokers more than beneficial owners. The question is not whether every affected person can participate directly. They cannot. The question is whether the institution understands which intermediaries are visible and which affected interests they omit.
An ISP does not automatically represent its customers' view on address transferability. A broker does not automatically represent the operator's view on registry accuracy. A civil-society group does not automatically represent end users' dependence on cheap connectivity. A government ministry does not automatically represent private networks. A regional-development organisation does not automatically represent small operators in every sub-region. A large resource holder does not automatically represent small holders. A small holder does not automatically represent future entrants. A registry does not automatically represent the region.
Lu Heng's public notes on representation make this point sharply, though from an interested position. They argue that a service region is an administrative category, not a sovereign people, and that no small group of meeting participants can honestly claim to speak for a continent or for end users as such. The argument should not be accepted uncritically; it is connected to a broader asset-rights and registry-limitation agenda. But the warning is sound: institutional labels can inflate participation into mandate.
AFRINIC's official regional design faces the same temptation in reverse. Because the board has regional seats and because the policy process is open to the internet community, it is easy to speak of "the African community" as if a coherent political subject has acted. Sometimes that shorthand is harmless. In a high-stakes scarcity dispute it becomes risky. A rule may be supported by visible African institutions and still impose costs on absent African operators. A rule may be opposed by commercial address holders and still raise genuine concerns for small networks. The geography of the speaker does not settle the economics of the rule.
The June 2025 election controversy also shows how the affected operator can disappear behind representation machinery. If a resource holder's voting authority is claimed by someone else, the issue is not only fraud in the narrow sense. It is the substitution of a visible participant for the affected member. Even a valid proxy should be understood as a second-best mechanism, not as proof that the member had equal voice. A governance design that relies heavily on proxies should be especially careful about confirmation, revocation, disclosure and audit.
The deeper point is that representation must be measured by exposure, not just participation. Who bears the cost if a transfer is delayed? Who bears the cost if a registry record is frozen? Who bears the cost if a board election fails? Who bears the cost if ICANN appoints an emergency arrangement? Who bears the cost if legal uncertainty raises the risk premium on AFRINIC-administered resources? The people bearing those costs should be mapped even when they do not speak.
An institution that does not map exposure will overlearn from the visible.
Scarcity makes representation more valuable
IPv4 scarcity changed the economics of participation. AFRINIC came late to the regional registry system and for a period retained a larger pool of unallocated IPv4 addresses than other regions. Internet Governance Project has described the arbitrage created by low administrative allocation fees and high secondary-market values. AFRINIC's exhaustion materials show the move into soft-landing phases, need-based rationing and small final allocations. The exact market price of an IPv4 address moves over time. The structural fact is stable: scarce IPv4 has economic value beyond the registry fee.
When a resource becomes valuable, representation around its rules becomes valuable too. A transfer policy is not merely a governance choice; it affects liquidity. A usage rule is not merely a stewardship claim; it affects business models. A documentation requirement is not merely diligence; it affects who can prove need cheaply. A board election is not merely institutional housekeeping; it affects enforcement posture, legal strategy and budget. A power-of-attorney rule is not merely meeting administration; it affects control over the institution that controls recognition.
This is why participation costs have distributional consequences. The actor able to shape a rule before others notice can influence scarcity rents. They may not receive the value directly, but the rule can raise or lower the cost of acquiring, leasing, transferring, defending or using address space. In an abundant environment, the effect of being absent is diluted. In a scarcity environment, absence can be priced.
AFRINIC's 2026 policy environment shows this clearly. Reporting by The Register in March 2026 described a new policy that in many circumstances prevents members from transferring AFRINIC-assigned IPv4 resources outside the region. Supporters can frame such a rule as regional stewardship. Critics can frame it as capital control or lock-in. The representation question is: whose costs were visible when the rule was shaped? Were small African operators who may need liquidity heard? Were future entrants who need addresses heard? Were customers dependent on cross-border services heard? Were holders with legitimate global customer bases heard? Were national operators worried about export heard? Were brokers and large holders overrepresented? Were public-sector networks considered? The legitimacy of the rule depends partly on those answers.
Scarcity also increases the incentive to influence elections. If a board can affect the institution's posture toward transfers, enforcement, litigation, bylaws and ICANN, then board seats carry economic significance. This does not mean candidates are corrupt or voters are bought. It means the cost-benefit calculation of participation changes. Organised campaigns become rational. Endorsement slates become more valuable. Proxy collection becomes more tempting. Legal challenges become more worthwhile. Public narratives become more strategic.
The danger is that institutions respond to scarcity by moralising participation rather than analysing it. One side accuses another of profiteering. Another accuses the first of gatekeeping. A third claims to defend the community. A fourth claims to defend continuity. These labels may contain truth, but they do not measure representation. The better question is how scarcity changes who can afford to appear and who has the strongest incentive to do so.
In a scarce-resource registry, representation should be treated like market infrastructure. It needs disclosure, audit trails and cost reduction. Otherwise the official process becomes the place where unequal participation costs are converted into apparently neutral rules.
What lower-cost representation would look like
A more representative AFRINIC would not try to turn every resource holder into a full-time policy participant. That would fail. The design goal should be lower-cost, higher-fidelity representation: members should be able to understand consequential choices, express exposure, delegate safely when needed and verify that visible participants are not being treated as a complete sample.
The first requirement is an impact register for consequential proposals. AFRINIC should publish a plain-language map of affected groups: small LIRs, end-user members, IXPs, universities, public-sector networks, large address holders, transfer participants, IPv4-dependent access providers, hosting firms and other enterprise networks. The map should state likely documentation costs, transfer effects, implementation burdens, legal uncertainties and operational dependencies. It should change when the draft changes.
The second requirement is participation-cost disclosure. A consensus report should not merely summarise support and opposition. It should identify whether discussion occurred mainly on the list or in a meeting, whether remote participation was practical, which languages had useful summaries, whether small operators or public-sector networks were heard, whether the affected category was represented mostly by associations and whether decisive stages followed late changes. This is not quota politics. It is evidentiary humility.
The third requirement is a clean authority system. Every proxy or power of attorney should be confirmed through a separate member channel. Members should receive receipts showing who is authorised, for which meeting, for which vote and how to revoke authority. Proxy concentration should be disclosed in aggregate before voting closes. Candidate endorsements should state who issued them, what criteria were used and what interests the endorsing body has. Slates should not be banned. They should be legible.
The fourth requirement is low-risk feedback and post-implementation review. A small operator should be able to report that a proposal raises legal cost, threatens a customer contract, creates documentation burdens or makes transfer planning harder without being forced into a factional public fight. After implementation, AFRINIC should publish privacy-protective data on processing times, documentation requests, denials, appeals, dispute holds and member categories affected. If a rule disproportionately burdens small operators or public institutions, the record should show it.
None of these reforms requires accepting an official story, a market story, an ICANN story or an NRS story. They require treating representation as infrastructure. The price of participation should be measured, reduced and disclosed.
Recovery will be judged by who can afford to matter
The board restoration announced after the annulled June 2025 election gave AFRINIC a chance to return to ordinary governance after years of paralysis. Reporting in early 2026 described improved staff morale, interim management appointments, work on a budget and action plan, and a strategy process for 2027-2030. Those are meaningful signs if they lead to predictable registry services, clearer authority and less litigation-driven uncertainty.
But recovery should not be judged only by whether the institution has a board, publishes a strategy, processes requests or wins support from other internet-governance bodies. Those are necessary. They are not sufficient. The representation test is whether ordinary members can afford to matter again.
That test has several watchpoints. First, member-authority verification. After the 2025 election controversy, AFRINIC must make direct authorisation boring. No member should discover at the point of voting that someone else claims to speak for them. No candidate should have to rely on rumours about proxy concentration. No receiver, board or committee should have to annul a vote because basic authority controls were unclear. If direct representation is costly, delegated representation must be exceptionally clean.
Second, policy readability. A restored board may be tempted to move quickly on backlog, bylaw reform, transfer policy, litigation response and operational repair. Speed is understandable, but it can raise participation costs. The more consequential the decision, the more important plain-language impact statements become. A policy that can be understood only by lawyers, market specialists and long-term insiders will reproduce the crisis in quieter form.
Third, meeting design. AFRINIC's return to in-person meetings across the region should be welcomed, but not romanticised. Travel again converts money into voice. Remote participation should be treated as a first-class channel. Meeting reports should acknowledge who was not in the room. Fellowships and travel support should be disclosed in aggregate so observers can see whether they broadened participation or merely reinforced familiar circles.
Fourth, legal-cost containment. Continuing litigation may be unavoidable, but governance should not become legible only through court filings. Members need clear separation between what a court order requires, what the board chooses, what staff implements, what policy says and what ICANN or NRO actors prefer. Without that separation, only legally advised participants can follow the institution.
Fifth, transfer and use policy incidence. AFRINIC's treatment of IPv4 transferability, regional use, leasing, documentation and abuse-contact obligations will reveal whether the restored institution understands the costs it imposes. If official channels are slow, discretionary or hard to interpret, activity will move into private arrangements and legal risk. If official channels are objective and auditable, more activity will remain in the registry record.
The central lesson is not anti-AFRINIC. A registry for Africa and the Indian Ocean remains necessary as a coordination layer. The lesson is anti-mythology. The existence of a regional registry does not prove regional representation. The existence of open meetings does not prove equal access. The existence of a board does not prove member voice. The existence of a slate does not prove capture. Each claim needs a cost account.
AFRINIC can become more legitimate if it makes participation cheaper, authority clearer and representation more auditable. It will become less legitimate if it treats the voices that can afford to appear as the voices of all those affected.
The conservative conclusion
The conservative conclusion is that representation in registry governance should be deflated, not inflated. AFRINIC should not claim to speak for a continent in any strong political sense. ICANN should not claim that global coordination automatically gives it superior democratic insight. NRS, LARUS or any market-oriented critic should not claim that resource-holder economics automatically represent end-user welfare. Smart Africa, operator associations and civil-society groups should not claim that organised regional endorsement is identical to member preference. Each can speak for its institution, members, evidence and interests. That is already valuable. It should not be overstated.
This deflation would make governance healthier. It would allow AFRINIC to say: we operate a registry, maintain records, process resources, run meetings, support policy development and administer member governance under defined constraints. It would allow critics to say: this rule imposes a cost on these holders and these customers. It would allow ICANN to say: this failure threatens continuity in these specific ways. It would allow associations to say: our members report these problems. It would allow courts to decide legal authority without being asked to validate every moral claim about the internet's future.
Participation costs will never disappear. The African internet is too large, multilingual, uneven and commercially diverse for a frictionless governance model. Some people will always be more present than others. Some organisations will always have larger budgets, better lawyers, better English, more experienced speakers and stronger incentives. The question is whether the institution treats those inequalities as noise or as part of the system it must manage.
AFRINIC's crisis shows what happens when representation costs are underdesigned. Resource governance moves into courts. Proxy authority becomes explosive. Endorsement slates carry heavy weight. Ordinary members struggle to know which claims matter. The visible community becomes both more important and less clearly representative. Scarcity makes every weakness expensive.
AFRINIC is a test case because its difficulties are unusually visible, not because they are unique. Every regional registry relies on a small active class to sustain open processes. Every registry faces the temptation to equate that class with the community. Every registry faces higher stakes after IPv4 scarcity. AFRINIC reveals the cost of the fiction sooner and more dramatically.
The room is not the region. The list is not the membership. The proxy is not always the member. The frequent speaker is not always the affected operator. The institution that recognises these limits can still govern. The institution that forgets them will find that representation, like IPv4, becomes scarce precisely when it is most needed.