The engineer understands the harm before he can describe it. His company is a modest access provider in a Francophone market. It buys equipment in English, files local paperwork in French, configures BGP with commands learned from vendor manuals, and answers customers in a mixture of French and local languages. A proposed AFRINIC policy is circulating on the Resource Policy Discussion list. The draft is written in plain enough English for him to see the danger. A phrase about transfer eligibility may reduce the future value of a block his company has treated as expansion reserve. A sentence about documentation may require corporate papers that local registries issue slowly. A reference to regional use may leave his lawyers unsure whether customers outside one jurisdiction will later be treated as evidence of non-compliance.
He does not lack intelligence, operating experience or stake. He lacks the recognised policy dialect. He can write that the proposal is bad for small providers. That will not be enough. To make a material objection he must identify the exact clause, distinguish allocation from assignment, explain whether the burden is technical, legal or economic, propose alternative wording, show that the harm is not merely private inconvenience, and do so in English before regular participants settle the frame. He may need to know whether "rough consensus" means visible assent, absence of sustained opposition, or a chair's judgment that objections have been answered. He may need to avoid sounding as if he opposes anti-abuse work, regional development or registry continuity. He may need to write for a list where old arguments are remembered by people he has never met.
The result is not silence in the simple sense. It is filtered evidence. The operator's knowledge exists, but the institutional record may never receive it in a form that survives. A fluent repeat participant can turn a weaker point into a durable objection. A less fluent operator can turn a stronger point into a comment that looks vague, late or local. In a consensus process, this difference matters. AFRINIC's policy system does not count votes on every clause. It reads arguments, weighs objections, tests consensus and produces a record that later supports board ratification, implementation, member confidence and, in a crisis, legal and market interpretation. Language therefore does not merely affect who feels included. It affects what the institution can know.
AFRINIC is the sharpest test because its language problem sits inside a larger institutional stress. The African Network Information Centre is the regional internet registry for Africa and parts of the Indian Ocean. It administers IPv4 addresses, IPv6 addresses and autonomous system numbers. Its records support WHOIS and RDAP, reverse DNS, routing-security functions, transfer confidence, member standing, billing, abuse handling and operational continuity. The region it serves contains Anglophone, Francophone, Arabic-speaking, Portuguese-speaking and many local-language operating communities. Its public materials may appear in more than one language, but the English text has special authority and the policy manual states that communication with AFRINIC is in English.
That choice has administrative logic. A registry needs authoritative text. The internet's technical vocabulary is heavily Anglophone. Staff cannot run every contractual, policy, legal and operational interaction in every language of the region without cost and risk. But a single official language also creates a fixed cost. It decides who can act quickly, who needs translation, who can draft, who can object, who can challenge a summary, and who can create the evidence that later proves a policy was or was not accepted. In ordinary settings that may look like inconvenience. In a scarce IPv4 environment, it becomes institutional economics.
AFRINIC's recent history makes the point harder to avoid. The dispute involving Cloud Innovation, litigation in Mauritius, years without ordinary board continuity, receivership, the annulled June 2025 election, reported controversy around powers of attorney and voter documentation, ICANN questions about proxy and election integrity, later efforts to restore a board, and continuing arguments over transfer policy all have one common feature: procedure itself became economically meaningful. When the registry's authority is trusted, policy language is a background rule. When trust is thin, every word can become a pricing factor. Buyers, lenders, lessors, operators, courts and members ask not only what the rule says, but whether the process that produced it heard the right evidence.
The core thesis is narrow. AFRINIC is a test case for how English-dominant process can alter representation, objection quality, consensus legibility and legal certainty in scarce-number governance. The issue is not whether English should disappear from registry operations. It will not, and in many contexts it cannot. The issue is whether a policy system that depends on English can recognise the evidentiary losses it creates and design around them. If it cannot, language fluency becomes a hidden qualification for shaping rules over scarce operational capital.
The first cost is translation into policy dialect
The most important translation in AFRINIC policy is not always between French, Arabic, Portuguese and English. It is between operating harm and recognised policy speech. An operator may know what a clause will do to customers, working capital, documentation burden, transfer optionality or routing continuity. The process hears that knowledge only after it has been converted into the dialect of the Policy Development Process.
That dialect has several layers. It uses technical terms such as allocation, assignment, sub-allocation, utilisation, reverse delegation, RPKI, ASN eligibility, abuse contact and registration data. It uses governance terms such as rough consensus, last call, board ratification, appeal, chair determination and community-developed policy. It uses legal-administrative terms such as member standing, authority, power of attorney, contractual obligation, compliance, revocation, transfer approval and dispute. It uses moral terms such as stewardship, fairness, regional need, conservation, stability and continuity. A useful objection often has to move through all four.
The cost is fixed before participation begins. Learning the English vocabulary of AFRINIC policy, understanding old list debates, knowing how chairs frame objections, recognising which documents are authoritative and writing in a way regulars treat as serious all require upfront investment. Large operators, consultants, lawyers, brokers and repeat participants can amortise that investment. A small ISP may have one or two people who handle routing, billing, procurement, abuse mail, customer escalations and registry paperwork. The same language requirement falls on both, but it is not the same economic event.
The cost also varies by task. Reading is cheaper than writing. Writing is cheaper than debating live. Drafting policy text is harder than commenting. Challenging a chair summary is harder than asking a question. Appealing a decision is harder than disagreeing in a meeting. A participant may be comfortable reading vendor documentation and still avoid public policy writing. Another may speak conversational English but not want to challenge a native speaker, a lawyer or a long-time community figure in a permanent archive.
Timing makes the fixed cost regressive. A proposal's first public framing matters disproportionately. The initial version defines the problem. Early objections define the legitimate issues. The first chair or community summary tells later readers what has been answered. Participants who must wait for translation, internal explanation or help from a bilingual colleague enter the discussion after the price has risen. By then, their objection may appear late, duplicative or resistant to closure. The same argument would have carried more weight if made fluently in the first week.
This is a high standard even for a native English speaker. For a non-native participant it is a compound cost. A French-speaking regulator may understand the public-law risk of a rule but not the mailing-list convention for saying it. A Portuguese-speaking operator may know that a documentation requirement is unrealistic for smaller corporate structures but write the point as a complaint rather than as an implementation objection. An Arabic-speaking lawyer may read the English text but not share the technical assumptions of the engineers who wrote it. An Anglophone engineer may read the technical text but lack the legal vocabulary needed to explain why "approval" creates discretion that markets will price.
The policy system then mistakes form for substance. A polished objection looks material because it identifies text, explains consequence and proposes a remedy. A rough objection looks weak because it lacks exact phrasing. Yet the rough objection may contain better field evidence. It may say that a local company registry cannot issue documents within the proposed deadline. It may say that a government network cannot disclose a customer relationship publicly. It may say that a small ISP's transfer financing will fail if registry confirmation takes an uncertain number of months. It may say that the local language version of a notice did not make clear that existing resources would be affected. These are not rhetorical defects. They are facts about implementation.
Consensus systems are especially vulnerable to this bias because they weigh the quality of objections rather than merely counting opponents. In a vote, weakly expressed opposition can still be recorded as a vote. In a rough-consensus process, the chair and participants must decide whether an objection has substance, whether it has been answered, and whether remaining opposition blocks progress. Language fluency becomes part of that judgment even when no one intends discrimination.
This does not mean every vague objection should block policy. A registry cannot run on unstructured anxiety. Serious objections should identify mechanisms and costs. But if the only participants who can express mechanism and cost are those already fluent in the recognised English policy dialect, the record will systematically underweight less fluent operating communities. The solution is not to lower the intellectual standard. It is to lower the translation cost of meeting it.
AFRINIC's language challenge is therefore more than multilingual access. It is an evidence-production challenge. The institution must ask whether it is hearing operational evidence or merely hearing the subset of evidence that can be converted into the expected idiom before the window closes. This is why "English is widely used in internet engineering" is not a sufficient answer. The relevant question is not whether many operators can decode English. It is whether they can act in English at the right time, in the right register, with the right procedural confidence.
Technical English and legal English make a second barrier
Language barriers in AFRINIC policy are not simply national or cultural. They are also professional. A technically fluent engineer and an English-speaking lawyer can misunderstand each other inside the same sentence. That matters because scarce-number policy now sits between routing, contract, corporate authority, insolvency, transfer markets and public claims about regional development.
Consider the word "use". In ordinary English it may mean that addresses are deployed on a network. In technical policy it may involve assignment, utilisation ratios, customer documentation, routing visibility, aggregation, conservation and hostmaster review. In legal argument it may involve contractual promises, representations in an application, implied limits, breach, reliance and remedies. In market analysis it may involve leasing, customer geography, beneficial control, option value and liquidity. A policy sentence about "proper use" or "regional use" therefore carries several meanings at once.
The same is true of "transfer". To an engineer, it may mean updating registry records so routing and contact information reflect changed control. To a lawyer, it may raise assignment of rights, corporate authority, warranties, sanctions, debt, insolvency and fraud. To a market participant, it means liquidity, price discovery, settlement risk and optionality. To a regional-development advocate, it may mean whether scarce addresses stay available to networks in Africa. A participant who objects in only one register can be dismissed by another.
"Consensus" has the same problem. In the internet tradition, rough consensus is not a poll. It asks whether material objections have been addressed and whether the group can move forward. In a legally stressed membership body, however, consensus may be read by outsiders as consent, mandate or evidence that affected parties had a chance to object. A non-native participant may not know how to distinguish "I do not like this" from "this remains a material objection because the draft gives staff unbounded discretion over existing holders." The latter survives. The former may not.
The result is a second dialect: technical-legal English. It is not the same as English as a foreign language. It is the language in which operational facts become institutional evidence. This dialect favours repeat participants who understand both the technical vocabulary and the governance consequences. It disadvantages those who know only their own side of the problem.
AFRINIC's crisis has made the dialect more demanding. A policy participant may now need to understand the Cloud Innovation dispute without making unproven claims, the receivership without treating it as ordinary management, the 2025 election controversy without asserting fraud beyond the reported record, ICANN's continuity concerns without accepting ICANN's framing as conclusive, and transfer policy without collapsing every debate into asset ownership or anti-market rhetoric. That is a sophisticated writing task.
Official materials can help by describing formal steps. They cannot solve the dialect problem by themselves. A manual that states the process in English is not the same as a guide that translates policy effect into operating examples across language communities and professional roles. A serious language architecture would produce not only translated text, but translated consequences: what the clause means for a small LIR, an end-user assignment, a merger, a public institution, a data centre, a hosting provider, a cross-border customer base, a leased block, a contact failure and a disputed authority file.
The risk is that AFRINIC hears mostly from those who can speak the second dialect. These people may be useful. They may also be unrepresentative. A policy archive dominated by technical-legal English can look rigorous while missing the knowledge of operators who cannot write in that form.
Translation delay gives the first mover the frame
Translation is often treated as a service: a page appears in another language, a meeting offers interpretation, or a summary is circulated after the fact. In policy economics, translation is also a timing device. The participant who reads the first authoritative version early can frame the problem before others receive an equally usable account. The delay need not be long to matter.
Policy debate has path dependence. The first draft names the problem. The first supporters supply moral language. The first critics identify accepted categories of objection. The first staff or chair summary condenses the issues. Later participants must enter a conversation already shaped by those choices. If the earliest and most precise conversation occurs in English, then English-speaking participants receive a framing option that others receive later, weaker or in summary form.
This timing effect is visible in several policy stages. At the proposal stage, English-first drafting gives the author control over problem definition. At the mailing-list stage, fast English replies create social proof that certain concerns are live and others are not. At the meeting stage, live English debate can set the mood before written translations or local explanations spread. At last call, late translation can turn first understanding into a procedural disadvantage. At implementation, English-only or English-prior guidance can decide who adapts early and who discovers the burden through a support ticket.
Translation delay is especially costly for objections. Support can be general and still useful. A participant can write that a proposal improves stewardship or stability. Objection requires precision. It must identify the defect. It must often propose narrower wording. It must show why the harm is material. It must do so before closure becomes the dominant institutional goal. A delayed objector therefore bears a heavier burden than a delayed supporter.
The same asymmetry applies to legal certainty. If English text is official and translated text is secondary, non-English readers may hesitate to rely on the translation when making business decisions. A French-language explanation may tell a holder that a transfer path exists, but counsel may still insist on reading the English original. An Arabic-speaking public body may translate the rule internally, then worry that the English phrase "may approve" or "shall approve" carries a discretionary nuance not captured in local language. A Portuguese-speaking operator may ask whether an example in translated guidance is binding or merely illustrative. Each uncertainty adds cost.
Every multilingual institution needs an authoritative text. Without one, disagreements over translation can paralyse administration. Public AFRINIC materials indicate an English-first hierarchy, and the policy manual states that communication with AFRINIC is in English. The difficulty is that an official-language hierarchy allocates legal certainty unevenly. The participant working in the official language can quote the operative text directly. The participant working through translation must decide whether nuance has shifted and whether a later dispute will be judged against the English original.
The hierarchy can also improve drafting if the institution lets it. Translation problems often reveal substantive ambiguity. If a clause cannot be translated without choosing between meanings, the original clause is probably not clear enough. A term such as "approval", "proper use" or "regional resource" may seem workable in English but force different legal choices in French, Arabic or Portuguese. The translation process can therefore be an early warning system for policy risk, not merely a publication chore.
Accurate translation of technical and legal material takes time, but policy timelines should be designed around that fact. Consequential proposals should not begin their real evidentiary clock until plain-language summaries, key definitions, and affected-party examples are available in the main operating languages. Last call should not close before non-English communities have had a genuine chance to read the current draft, not merely an earlier version. Chair summaries should identify whether translation lag affected participation.
This is not ceremonial inclusion. It is record quality. A consensus record that develops before affected language communities can understand the practical consequence is a weak record. In a post-crisis AFRINIC, weak records become expensive because every policy can later be read through distrust. Legal certainty should not become a private good purchased by English-fluent actors; it should be a public feature of the registry record.
Policy wording moves value after IPv4 scarcity
Language matters more after scarcity because policy wording moves value. Before IPv4 exhaustion, registry language mostly governed entry into a managed pool. After exhaustion, many clauses affect resources already embedded in networks, customer contracts, financing plans, security services and transfer markets. The same sentence can change liquidity, bargaining power and dispute risk.
A phrase about "regional resources" can affect whether a block may move out of the AFRINIC region. A requirement for "written approval" can turn record updating into a discretionary checkpoint. A term such as "proper utilisation" can invite review of a holder's business model if not bounded. A rule about "abuse contact" can remain directory hygiene or become an enforcement hook depending on remedies. A definition of "assignment" or "sub-allocation" can determine whether downstream customer use is transparent or risky. A last-call summary saying objections were "addressed" can become later evidence that the economic burden was accepted.
This is why policy language should be read as market infrastructure. It does not merely express values. It structures transactions. A clear, narrow, objective rule lowers transaction costs. A broad, moralised or ambiguous rule raises them. It may give the registry flexibility, but flexibility held by a low-liability institution becomes risk held by members and markets.
AFRINIC's transfer-policy debates show the stakes. Public reporting and circulated draft material describe controversy over whether AFRINIC-issued IPv4 resources can move outside the region, how regional classification should work, and whether transfer restrictions protect African networks or trap address value. Supporters of tighter rules may see regional stewardship. Critics may see capital control, lock-in or a registry asserting authority beyond narrow recordkeeping. The language chosen decides which argument sounds normal.
In such debates, non-English operating communities may face a double disadvantage. First, they may need to understand how the wording affects value. Second, they must express the value effect without sounding as if they reject the public goods invoked by proponents. A Francophone operator that says "we need transfer flexibility" may be heard as private interest. A fluent policy regular can say "the proposed text converts a recording function into a discretionary approval regime that will reduce liquidity for existing holders without a proportional fraud-control benefit." The latter is the same economic point in recognised language.
The wording also shapes evidence quality. If a proposal uses broad terms such as stability, stewardship or regional benefit, objectors must supply concrete harms. If the proposal uses concrete operative terms, supporters must justify the chosen mechanism. Broad language therefore shifts the evidentiary burden to critics. In a multilingual environment, that burden falls hardest on those least able to produce polished English objections.
The answer is drafting discipline. High-consequence policies should avoid moral words unless they are tied to operational tests. If the risk is fraud, define fraud indicators. If the risk is duplicate recognition, define the evidence needed. If the risk is unreachable abuse contacts, define contactability and cure. If the risk is transfer speculation, explain the concrete registry harm rather than using speculation as a label. If the purpose is regional supply, identify how the rule increases supply rather than merely confines existing value.
Clear wording is not a concession to market actors. It is the minimum requirement for legitimacy where language fluency is uneven and value is scarce.
Objection quality is a language outcome
AFRINIC's process, like other RIR policy systems, depends on objections that can be evaluated. A chair cannot test consensus against every private concern. The process needs objections to be clear enough to answer. This is sensible. It also means that language quality becomes a component of institutional power.
A strong objection has several traits. It identifies the exact text. It explains the mechanism of harm. It distinguishes technical, legal, financial and implementation risk. It avoids irrelevant rhetoric. It proposes a remedy or narrower clause. It persists through revised drafts. It appears before closure. It is usually written in confident English. Each trait is partly substantive and partly linguistic.
Non-native or non-specialist objections often fail in predictable ways. They may state the conclusion without the mechanism. They may use ordinary words where policy expects defined terms. They may sound emotional because the writer lacks procedural phrasing. They may omit alternative wording. They may repeat points already addressed because the writer did not follow earlier debate. They may appear late because internal translation took time. They may be too short because the author fears making a mistake in public. The process then classifies them as weak.
Some weak objections are weak. But many are weakly translated. A line such as "this policy will stop our business" may hide a precise claim: the transfer condition will prevent a financed acquisition from closing because the buyer's customers are outside the region, even though routing and abuse accountability would remain clear. A line such as "small companies cannot do these papers" may hide a documentation-cost claim: local corporate registries, notarisation systems and public-sector approvals cannot meet the proposed deadline, so large firms with counsel will comply while smaller holders face delays. A line such as "we do not understand the meaning" may hide a drafting defect.
The institution should treat such comments as leads for evidence, not as noise. Chairs and staff can ask structured follow-up questions: Which clause creates the difficulty? Is the concern about existing resources or future allocations? Is the burden translation, documentation, cost, timing, legal authority or operational continuity? Would an example help? Can the concern be summarised without naming the company? Such questions turn weakly expressed experience into usable evidence.
This matters because AFRINIC's high-stakes record is already contested. If the archive shows only polished arguments from repeat participants and thin comments from occasional operators, later readers may infer that the polished class carried the serious evidence. That may be false. The serious evidence may have existed in less fluent form and been lost at intake.
Objection quality also affects appeals. A participant who wants to challenge a chair decision must usually show that a material issue was not addressed. That requires procedural literacy and language precision. A non-English operating community may therefore be disadvantaged twice: first when objecting, and again when challenging the treatment of the objection. The appeal path may exist formally while remaining expensive in practice.
The remedy is not to accept every objection at face value. It is to create an objection translation layer. High-consequence proposals should provide structured objection templates in the main operating languages: clause, affected party, existing or future resource, operational effect, legal effect, documentation effect, time effect, proposed fix, confidentiality concern. Submissions can then be translated into the official record with the original preserved. A chair report can quote the English synthesis while noting the language and affected category of the original evidence.
Such a system would make objections more comparable. It would also reduce the advantage of those who know how to write in the expected policy idiom. The point is not politeness. It is better evidence.
Summaries, minutes and consensus legibility filter dissent
The policy record is not only the raw exchange. It is also the summary: meeting minutes, chair assessments, last-call notices, board papers, implementation notes, public announcements and later explanations. In a multilingual system, summary is dangerous because it can become a second filter. The first filter is whether an affected person can express the concern. The second is whether the institution summarises it faithfully.
Summaries always simplify. A meeting cannot reproduce every sentence. A chair report must distinguish material issues from repetition. A board paper must be usable. The problem arises when summary language strips away the economic or linguistic context of an objection. A comment from a Portuguese-speaking small operator may become "concern about documentation burden". That may be accurate but incomplete. Was the burden tied to public-sector notarisation? To local corporate registry delays? To translation of legal documents? To the fact that large operators can buy counsel while smaller ones cannot? The policy relevance lies in the detail.
Summary bias also affects late or translated comments. If an objection arrives after English debate has shaped the issue, a summary may treat it as confirming an already known category rather than as evidence that a language community reached understanding late. That distinction matters. A late comment can be strategic delay. It can also reveal that notice failed. The record should say which appears to be the case.
In AFRINIC's crisis setting, summary bias becomes part of legal and market risk. Later disputes may ask whether a policy had consensus, whether objections were material, whether members understood the effect, whether the board acted on a reliable record, or whether implementation exceeded the policy. If summaries understate language-driven evidence loss, the institution's proof is weaker than it looks.
The same issue applies to election and member-meeting records. If a member reports a proxy concern, credential issue or authority dispute in a language other than English, how is that classified? Is it an administrative ticket, a voting irregularity, a legal claim, or a help request? If a member cannot phrase the problem in English at the decisive moment, does the record show a weaker incident? AFRINIC's annulled June 2025 election shows why authority evidence must be precise. The public record of reported power-of-attorney controversies, ICANN questions and receiver action does not prove every allegation, but it proves that record quality determines trust.
Minutes also shape memory. Repeat participants can later cite the summary as if it were the whole debate. Newcomers rely on it because they cannot read every archived message. If summaries consistently compress non-English operational concerns, the next policy cycle begins with a distorted institutional memory. The process then compounds language bias over time.
A stronger summary discipline would classify evidence by language channel, affected category and unresolved mechanism. It would distinguish questions from objections, objections from implementation examples, and implementation examples from general sentiment. It would state whether translation or interpretation was available, whether any substantive comments arrived through translated channels, and whether those comments changed the draft or chair reasoning. Where the chair rejects an objection, the summary should explain the reason in plain language, not merely say that the objection was addressed.
This may sound bureaucratic. It is cheaper than litigation over what the record meant. In AFRINIC's environment, the cost of a careful summary is small compared with the cost of another legitimacy dispute.
Consensus is not only reached. It is made legible. People must be able to see why the process closed, which objections mattered, which did not, and what the outcome means for future behaviour. In a scarce-number registry, a losing participant may still comply if the reasons are clear. A resource holder may accept a transfer rule it dislikes if it can see the fraud-control rationale, the limits of discretion and the appeal path. A public institution may accept timing rules if official summaries explain the legal authority. If the reasoning remains locked in English policy dialect, losers will suspect that consensus was declared inside a club.
The language of community can worsen this. AFRINIC, like other RIR institutions, often relies on bottom-up and community language. Such language is not empty. Open process is a real asset. But "the community" becomes hard to defend when large parts of the region experience the operative record as foreign-language expert discussion. A Francophone, Arabic-speaking or Portuguese-speaking member may see public materials, but not feel that the decisive reasoning was made accessible before the outcome hardened.
Legibility is also distributional. Large actors can translate the record privately. They can hire counsel to read English minutes, compare versions and advise management. Small actors depend on public explanations. If the public record is difficult, private translation becomes another form of advantage. Low participation then makes future consensus thinner. Thin consensus makes policies easier to challenge. Challenges make governance more legalistic. Legalism increases the language burden. The cycle is self-reinforcing.
Breaking the cycle requires treating consensus explanations as public goods. AFRINIC should publish for consequential policies a short official effect statement in the main operating languages: what changed, who is affected, what objections remained, why they did not block adoption, what actions members should take, what services are unaffected, and how to seek review. Such a statement would not replace the official English text. It would make the official outcome intelligible enough for compliance and confidence.
Multilingual communities do not experience one policy process
It is tempting to describe AFRINIC as a multilingual region and stop there. That phrase is too blunt. Different language communities do not merely translate the same process. They experience different process economies.
Anglophone participants often receive the official text first, can follow the main list more easily, and may be more comfortable in live debate. Francophone participants may have strong technical and regulatory communities but face the cost of converting local legal and commercial experience into English. Arabic-speaking participants may work across legal systems and public-sector institutions where formal authority and translation chains are different from the private-company assumptions embedded in policy examples. Portuguese-speaking participants may be fewer in number but tied to important Lusophone African markets where local corporate and telecom realities do not match English templates. Local-language operators may depend on intermediaries almost completely.
These communities also differ internally. A large telecom group in a Francophone country may have better English policy capacity than a small Anglophone access provider. A global company operating in an Arabic-speaking market may have multilingual counsel, while a domestic network does not. A regulator may have translation capacity but limited technical-policy familiarity. An engineer may be multilingual but not authorised to speak for the organisation. The point is not to rank languages. It is to map the costs of using the policy process.
The cost also varies by issue. A straightforward technical update may be easy to translate. A transfer policy is harder because it mixes market value, legal authority, regional rhetoric and registry discretion. An abuse-contact rule may look simple until remedies are discussed. A board-election rule may require knowledge of corporate authority, powers of attorney, registered member status, proxy concentration and voting receipts. A policy affecting RPKI or reverse DNS may require technical precision that cannot be safely summarised in broad language.
Language communities also receive informal information differently. Policy meaning is not formed only in official texts. It forms through meeting side conversations, private explanations, mailing-list tone, association briefings, national operator groups, lawyers, consultants, journalists and campaign materials. English-speaking insiders may hear nuance that a translated summary never conveys. Francophone or Arabic-speaking operators may rely on local networks that interpret the policy through national concerns. Portuguese-speaking participants may have fewer peers in the central debate and therefore weaker informal correction. These differences shape confidence.
AFRINIC should not try to solve this by imagining one perfectly equal multilingual public sphere. That is unrealistic. It should instead design for known asymmetry. Where a proposal affects existing resources, it should identify which language communities are likely affected and what outreach reached them. Where comments come mainly from one language community, chair reports should say so. Where translations were unavailable or late, the consensus weight should be adjusted. Where a policy relies on legal terms, translation notes should identify risk points.
The point is evidentiary humility. A policy process can be open and still have uneven language reach. The record should show that unevenness rather than hide it under the word community.
Language interacts with litigation, not just participation
Language barriers become sharper when the institution is litigated. In a calm environment, unclear policy language may be corrected in the next cycle. In a crisis, unclear language becomes evidence for legal claims, public accusations and market discounts. AFRINIC's history shows this clearly.
The Cloud Innovation dispute turned resource-use language, contractual obligations, registry authority and proportional remedies into matters of litigation. Public analysis from the Internet Governance Project treated the conflict as a political-economy dispute over scarcity and registry power, not merely a technical compliance argument. AFRINIC and its supporters have described litigation and procedural roadblocks as threats to registry recovery. Lu Heng, NRS and Larus-linked commentary have argued that registry discretion over economically critical resources lacks matching liability and review. The public record is adversarial, and not every claim should be accepted. But the existence of the dispute shows that wording is not harmless.
In such a setting, language barriers affect legal exposure. A member may avoid public comment because a poorly worded English statement could be quoted in litigation or factional debate. A company may refuse to object unless counsel approves the language. A regulator may not want a technical employee's English email to be treated as an official national position. A small operator may fear that an imprecise statement will align it with one side of a dispute. Silence under these conditions is rational, not necessarily consenting.
Legalised environments also reward English-proficient professionals. Lawyers, policy consultants and repeat institutional actors become more valuable because they can phrase positions safely. Ordinary operators withdraw or delegate. The record then becomes more professional and less representative. It may look stronger because the prose is more polished. It may actually be weaker because operating evidence has been chilled.
Translation in litigation also affects member confidence. If a policy debate produces English summaries that courts, ICANN, AFRINIC or critics later cite, non-English members may wonder whether their concerns were ever part of the legally relevant record. If a court order or receiver communication is understood mainly through English-language reporting, member perception varies by who translates it locally. If ICANN asks questions about proxy rules or election integrity in English, members in other language communities may receive the controversy as filtered narrative rather than as a precise procedural issue.
This does not mean AFRINIC should conduct litigation politics in multiple languages. It means governance design should avoid adding avoidable language ambiguity to an already legalised environment. Election authority forms, proxy notices, member receipts, transfer conditions, appeal windows and implementation consequences should be written in plain English and supported by reliable translations or effect summaries. The more legal risk a document carries, the less it should depend on insider interpretation.
The same caution applies to public accusations. The 2025 election controversy involved reported allegations over powers of attorney and voter documentation. A careful public analysis should not assert unproved misconduct. But the episode shows that when authority documents, member communications and challenge procedures are not universally legible, every side can claim the record supports its interpretation. Language clarity is therefore part of election integrity.
Litigation raises the cost of bad translation. A mistranslated or poorly understood policy may not merely confuse a member. It may become a dispute over consent, authority or reliance. AFRINIC's recovery should treat language risk as legal risk.
Language decides whose evidence is admissible to the institution
Evidence in number-resource policy is not only documents. It includes operating examples, timing burdens, customer dependencies, local corporate-law realities, payment and notarisation frictions, public-sector authority chains, national regulatory constraints, market behaviour, and the practical consequences of delay. Language determines how much of that evidence becomes admissible to the institution.
The word "admissible" is useful even outside court. A policy process receives many statements. Some are treated as relevant evidence; others are treated as opinion, anecdote, confusion or noise. The standard for admission is partly informal. Does the statement use recognised terms? Does it answer the proposal? Does it fit the stage of debate? Is it written by someone known? Is it easy to summarise? Language fluency affects each question.
An English-speaking repeat participant can convert anecdote into evidence. They can write: "The implementation burden will fall unevenly because smaller holders in jurisdictions with slower corporate registries cannot obtain updated authority documents within the proposed cure period; the draft should include a ninety-day cure and preserve RPKI and reverse DNS during review." A less fluent operator may write: "Our papers take long time, do not stop our network." The first sounds like policy. The second sounds like concern. They may describe the same fact.
This matters for scarce-number governance because the most important evidence is often local and messy. AFRINIC's region includes multiple legal systems, different company registries, state-owned networks, small family-owned access providers, universities, IXPs, mobile operators, hosting firms, cross-border customers and public agencies. A rule that looks clean from the policy desk may work differently in each environment. If local evidence must pass through English technical-legal filters before it is recognised, the policy record will overrepresent environments where those filters are cheap.
Language also affects quantitative evidence. A staff impact report may ask for data: number of affected resources, processing times, documentation failures, transfer delays, abuse-contact cure rates. Operators who cannot easily explain their situation in English may not submit the data. Their cases then remain invisible. The absence of data is interpreted as absence of harm. This is a classic evidence trap.
The trap can be avoided by structured intake. AFRINIC could ask for specific evidence in multiple languages: What resource type is affected? What country? What document is hard to obtain? What typical timeline? What service would be affected? What customer category? What alternative wording would reduce harm? Could the evidence be published anonymously? This makes evidence easier to provide and easier to compare.
The institution should also separate evidence from advocacy. A small operator should be able to submit an implementation fact without endorsing a faction in AFRINIC's wider dispute. A regulator should be able to explain national paperwork constraints without taking a position on Cloud Innovation, NRS, ICANN or a board slate. Language templates can help by focusing submissions on mechanism rather than allegiance.
The long-run benefit is a better archive. Future policy participants would not have to infer local reality from scattered comments. They could see structured evidence from language communities and affected categories. That would reduce the power of repeat participants who claim to speak for absent operators. It would also reduce the temptation for official actors to equate lack of polished English objection with lack of harm.
The market prices language uncertainty
Language uncertainty eventually reaches markets. A registry rule that is hard to understand in the official language, harder to translate, and harder to explain to counterparties becomes a transaction cost. It does not matter whether the uncertainty began as a translation problem or a policy problem. The buyer, lender, customer or lawyer sees risk.
In a transfer, parties ask whether the block can move, what evidence is needed, how long approval will take, whether existing use creates issues, whether RPKI and reverse DNS remain stable, whether a dispute will be flagged, and whether later policy reinterpretation could affect the transaction. If the answer depends on specialised English policy history, parties demand protection. They add warranties, conditions, escrow, longer timelines, indemnities or discounts. A non-English holder may pay more for counsel merely to make the transaction legible.
In a leasing or customer-use arrangement, language uncertainty can push activity away from transparent registry records. If holders do not know how AFRINIC will describe downstream use, or if policy language around assignments and sub-allocations is unclear across languages, they may choose private arrangements that preserve optionality but reduce public visibility. That is bad for registry accuracy. A registry that wants clean records should make the official path linguistically cheap.
In compliance, language uncertainty can create overreaction. A member receiving an English notice may not know whether it is a routine data-quality request, a warning, a contractual escalation or a potential service threat. A large operator asks counsel. A small operator panics or ignores it. Both responses are inefficient. Clear multilingual effect statements reduce false alarms and reduce missed deadlines.
In elections, language uncertainty weakens authority. A proxy form, power of attorney, candidate statement, voting instruction or member-status explanation must be understood by organisations operating under different languages and legal categories. If members are unsure what they are authorising, delegation becomes suspect. If challengers are unsure what documents count, every result becomes vulnerable to narrative attack. The June 2025 AFRINIC election controversy shows the cost of weak authority records. Language is not the only issue, but it is part of the proof chain.
Market confidence depends on boring predictability. AFRINIC-administered resources should not require special language risk analysis beyond ordinary diligence. A buyer should not wonder whether the French summary and the English manual imply different transfer effects. A public-sector network should not need to guess whether an English notice binds it before internal translation. A small ISP should not need a regional insider to know whether an objection is still timely. Each uncertainty adds to the governance discount attached to the registry.
This is why language spending should be treated as infrastructure spending. Translation, interpretation, plain-language summaries, examples, glossary maintenance and multilingual evidence intake are not public-relations extras. They reduce transaction costs. They improve record quality. They lower legal risk. They increase confidence that policies are understood before they bind scarce resources.
The cost is real. Good translation requires money and staff. But the alternative is not free. It is paid through disputes, delays, weak participation, poorer evidence, market discounts and avoidable suspicion.
A lower-cost language architecture
A serious AFRINIC language architecture would begin by classifying policy risk. Not every document needs the same treatment. Routine website updates, low-consequence technical notices and minor editorial changes may not justify full multilingual machinery. Policies affecting existing-holder rights, transfers, resource status, reverse DNS, RPKI, revocation exposure, member voting, proxy authority, fees or legal obligations should trigger a higher language standard.
The first element is a controlled glossary. Terms such as allocation, assignment, sub-allocation, utilisation, regional resource, transfer, merger, abuse contact, revocation, suspension, member standing, proxy, power of attorney, last call, rough consensus, ratification, appeal, RPKI, reverse DNS, RDAP and WHOIS should have stable explanations in English, French, Arabic and Portuguese. The glossary should explain not only literal meaning, but policy effect. Where no exact equivalent exists, the note should say so.
The second element is plain-language effect summaries. Each high-consequence proposal should have a short explanation in the main operating languages: what changes, who is affected, whether existing resources are covered, what evidence may be required, what services could be affected, what actions members should take, how to object, and what happens if the proposal is adopted. The summary should not advocate for the proposal. It should lower the cost of understanding it.
The third element is timing discipline. Comment periods should be calculated from the availability of current-language summaries, not merely from the posting of the English draft. If a proposal changes materially, the affected summaries should change too. Last call should not become a trap for participants who first receive usable translation after the social consensus has hardened.
The fourth element is structured objection intake. AFRINIC should provide templates that let participants submit clause-specific objections or implementation evidence in the main operating languages. The templates should ask for mechanism, affected category, resource type, country or legal context where relevant, operational effect, documentation effect, time effect and proposed remedy. Staff or chairs can translate the structured result into the official record while preserving the original.
The fifth element is multilingual chair reporting. Consensus reports should identify which language channels produced substantive comments, whether translation was available at key stages, whether late objections reflected late understanding, and how material objections were addressed. A report that says "no objections remained" should say whose objections were visible enough to assess.
The sixth element is implementation examples. For policies affecting transfers, contact validation, resource review, RPKI, reverse DNS, voting authority or documentation, AFRINIC should publish examples across common scenarios: small ISP, university, government network, merger, name change, cross-border customer, leased use, disputed authority, stale contact, public-sector signatory, and transfer with time-sensitive financing. Examples often translate policy better than prose.
The seventh element is confidential evidence channels with public aggregation. Some operators will not disclose commercial exposure in a public archive. They should be able to submit sensitive implementation facts confidentially, with AFRINIC publishing aggregated concerns and how they were considered. This prevents public-speaking risk from erasing evidence.
The eighth element is post-implementation review by language and member class. Did certain language communities file more support tickets? Did translation errors produce repeated misunderstandings? Did small operators miss deadlines? Did public institutions request more clarifications? Did transfer delays cluster around documentation from particular legal systems? The answers would turn language policy from aspiration into management data.
Such a system would not make AFRINIC perfectly multilingual. It would make language costs visible, measurable and lower. That is the realistic institutional target.
Source boundaries and uncertainty
The public record around AFRINIC is adversarial. A careful analysis should not treat every source as the same kind of evidence. AFRINIC's own materials are useful for formal facts: the registry's role, service region, policy process, language hierarchy, policy manuals and public notices. They should not be treated as conclusive framing for the economic legitimacy of that process. An institution's description of its own openness is an exhibit, not the verdict.
Independent reporting by The Register is useful for the chronology of boardlessness, receivership, attempted elections, the annulled June 2025 process, proxy and power-of-attorney controversy, later board formation, ICANN involvement, bylaw disputes and continuing litigation. Those reports do not prove every allegation made by every party. They establish that AFRINIC policy and governance operate under contested legitimacy, which is the relevant setting for language analysis.
The Internet Governance Project's analysis is useful because it treats the Cloud Innovation dispute as a political-economy conflict around scarcity, transfer markets and registry power. It is analysis, not a court judgment. KrebsOnSecurity's 2019 reporting on alleged address-record manipulation is relevant because it shows the economic stakes attached to registry records, not because it proves every later claim about current governance. Public notes by Lu Heng and materials from NRS, LARUS and related projects are indispensable for understanding the address-holder critique of registry discretion, liability asymmetry, insider process and continuity risk. They are also interested sources and should be attributed as such.
Official statements from ICANN, the NRO and peer registries are useful when they record continuity concerns, coordination obligations or procedural interventions. They should not be allowed to settle whether a particular AFRINIC policy was economically legitimate, linguistically accessible or proportionate. Continuity of the numbering system is a real concern. It does not erase the need to examine how policies are written and who can object.
Several facts remain uncertain or contested in the public materials reviewed for this analysis. The legal merits of AFRINIC's claims against Cloud Innovation and Cloud Innovation's claims against AFRINIC are not decided by this analysis. The validity of particular powers of attorney in the 2025 election controversy is not resolved here. The long-term effects of a restored board, Smart Africa-backed candidates, transfer-policy changes and global RIR lifecycle revisions remain to be observed. These uncertainties strengthen rather than weaken the language point. When facts are contested, the quality and accessibility of the record become more important.
The structural claim is more stable: in AFRINIC, English-dominant policy process can affect who frames issues, who objects effectively, how summaries describe dissent, how consensus becomes legible, how courts and markets read the record, and how members price the registry's authority. That claim does not require proving bad faith by any side. It follows from the economics of language in a high-stakes institution.
How to tell whether language reform is real
The test is timing first. For any policy affecting transfers, existing resources, voting authority, registry services or documentation burdens, current plain-language summaries should be available in French, Arabic and Portuguese while the debate is still live. If translations arrive after the frame has hardened, they are evidence of outreach, not equal participation. The same principle applies when a proposal changes materially: a translation of the obsolete draft does not make the current draft accessible.
The second test is how objections are treated. Chair reports should show whether comments from translated channels were treated as material when they identified real mechanisms. If those comments are consistently summarised as questions or concerns while polished English objections receive detailed answers, language bias remains. The record should also distinguish late obstruction from late understanding. A late objection by a repeat participant may be tactics. A late objection from an affected language community may reveal that the process did not reach that community early enough.
The third test is legal and market vocabulary. Key terms such as transfer, regional resource, approval, assignment, revocation, suspension, proxy, authority and appeal need consistent explanations across languages and examples. Proxy forms, power-of-attorney guidance, voting instructions, candidate materials and member-authority receipts should be plainly intelligible across operating languages. Any reform dealing with resource members, registered members, community resolutions, voting rights or corporate-law alignment must be translated as legal effect, not merely as words.
The fourth test is implementation data. After policy adoption, AFRINIC should measure whether support tickets, missed deadlines, incomplete forms or transfer delays cluster by language community, country or organisation type. If parties continue to require special counsel, indemnities, discounts or long diligence because policy language is hard to interpret, reform has not lowered market risk. The market will not issue a language report. It will price uncertainty quietly.
The final test is member confidence. More multilingual pages are not enough. The practical question is whether ordinary members believe they can understand a consequential policy, submit evidence, object, delegate authority, verify a vote and comply without hiring an English-speaking insider. A careful institution will say which participants, language channels and affected categories were heard. A weaker institution will say "the community" and let the phrase carry more weight than the record supports.
The conservative conclusion
AFRINIC does not need a romantic theory of multilingualism. It needs a cost account. English will remain central to internet coordination, technical standards, registry operations and many legal documents. A single authoritative text may be unavoidable for some purposes. But English dominance should be treated as an institutional cost, not as a neutral background fact.
The cost is paid in several ways. Non-English operators enter debates later. Objections lose force when operating harm is not translated into policy dialect. Summaries compress evidence that is already filtered. Legal certainty becomes easier for those who can rely on the official text directly. Markets add diligence and discounts where language makes rules ambiguous. Members lose confidence when consensus is legible only to the fluent. Courts and external institutions receive records that may look cleaner than the underlying participation.
This is not an argument that every policy outcome disliked by a non-English participant is illegitimate. Nor is it an argument that language explains all of AFRINIC's crisis. The institution's problems involve scarcity, litigation, board authority, record integrity, transfer markets, member trust, proxy controls, external intervention and competing institutional theories. Language sits inside those problems. It decides which facts are visible when decisions are made.
The strongest language reform would be boring: stable glossaries, timely summaries, structured objections, faithful minutes, implementation examples, authority notices, measured support outcomes and honest chair reports that describe who was heard and who was not. Such reforms would not guarantee agreement. They would improve the quality of disagreement. In AFRINIC's setting, that is valuable. A registry recovering from contested governance does not need theatrical unanimity. It needs records that reasonable losers can understand.
The small Francophone operator at the beginning should not have to become a policy lawyer to make a material point. The Arabic-speaking public network should not need a private interpreter to know whether a vote or proxy is valid. The Portuguese-speaking ISP should not discover a transfer cost only when English implementation guidance is applied. The Anglophone engineer should not be advantaged merely because the official dialect matches his training. The policy system should translate operating evidence into institutional knowledge before consensus is declared.
Language is often described as access. In scarce-number governance it is more than that. It is a production function for evidence, objections, certainty and trust. AFRINIC's test is whether it can keep English as an operating tool without letting English fluency become a hidden property right over the policy record.