Summary

  • ICANN's 2001 recognition criteria told a new regional registry how to enter the system, but they did not define a mature route for suspension, conditional recognition, emergency limitation, withdrawal, independent evidence testing, or cure after a recognized registry later fails.
  • The AFRINIC crisis made the gap visible. ICANN's 25 June and 3 July 2025 letters placed AFRINIC on notice that a compliance review might be necessary, identified ICP-2 obligations, and demanded evidence. The letters did not themselves initiate a completed review or define what sanctions, service limits, or remedies would follow.
  • A legitimate recognition power must be narrow enough to protect registry continuity without converting ICANN into a political supervisor. The missing rule should define triggers, evidence custody, notice, cure, interim safeguards, holder protections, independent review, publication, appeal, and handback before the next crisis forces improvisation.

Recognition is the hidden sanction

Recognition looks ceremonial until it is threatened. A regional internet registry does not merely carry a name. Recognition tells operators, governments, members, courts, lenders, cloud platforms, address buyers and other registries that one institution is the accepted service point for a large geographic region. It anchors number-resource allocations from IANA, registry services, public records, reverse-DNS delegation, RPKI trust relationships, transfer confidence, membership standing and the assumption that routine updates will be accepted by the rest of the system.

That makes recognition a powerful lever even before any formal withdrawal. When ICANN says that a registry may be reviewed for continued compliance, the market hears more than a governance concern. A buyer may delay an IPv4 transfer. A bank may discount address-dependent collateral. A public network may seek warranties. A court may wonder whether a local corporate remedy has wider technical consequences. Other registries may ask whether emergency cooperation will be required. A registry's board, receiver or senior staff may find that a letter from ICANN changes the bargaining field inside a domestic dispute.

This is why the recognition power cannot remain mostly implied. A power that can alter expectations across an entire region needs rules before it is used in anger. Those rules need not make ICANN a sovereign regulator. They should do the opposite. They should prevent every crisis letter from being treated as an open-ended political weapon by defining what ICANN may evaluate, what it may demand, what it cannot decide, what evidence counts, what cure looks like and how affected parties can challenge a conclusion.

The present historical record is uneven. The original recognition criteria were created to evaluate new regional registries. The ICANN-hosted ICP-2 text says the document was accepted by the ICANN Board on 4 June 2001 as essential requirements for recognizing new regional internet registries and as a framework for considering applications. It contains entry criteria about regional scale, community support, neutrality, technical capability, policy consistency, funding, record keeping and confidentiality.

Those are serious criteria, but entry criteria are not an exit rule. A founding review asks whether a candidate can begin. A crisis review asks what happens if an incumbent registry may no longer meet the duties that justified recognition. The first question is comparative and prospective. The second is remedial, disruptive and tied to existing reliance. Confusing them gives ICANN a large practical power with too little procedural discipline.

The difference matters because no regional registry starts as a blank file after recognition. It accumulates service agreements, member records, historic allocations, reverse-DNS zones, RPKI material, staff knowledge, bank accounts, court exposure, domestic corporate obligations and expectations from public agencies. Withdrawal or suspension cannot be treated as the reverse of approval. It is not enough to ask whether the registry still looks like the candidate described in 2001. The rule must ask what consequence follows a failure and how the public function survives while rights are tested.

Recognition is therefore the hidden sanction in number-resource governance. It is not a fine, a contract damages claim or a court judgment, but it can shift value because other actors rely on it. The sanction becomes harder to evaluate when it is not named as one. A letter can say it is only a notice. A registry can say it is political pressure. A resource holder can say it is a threat to continuity. A court can say it is an external institutional view. All four can be partly right if the underlying power is undefined.

The first reform is linguistic discipline. ICANN and the RIRs should distinguish recognition, audit, compliance review, warning, conditional remediation, temporary service support, emergency service transfer and withdrawal. These are different tools. They should not be invoked as a cloud of authority. Once the vocabulary is clear, the institution can design triggers and remedies that match the risk.

The original criteria created entry control, not exit control

ICP-2 was a product of the early regional registry era. It addressed a world in which APNIC, ARIN and RIPE NCC were already operating, while Africa and Latin America were moving toward their own regional registries. The text expected the number of registries to remain small because overlapping service regions could fragment address space, complicate coordination and confuse the community. It also recognized the practical difficulty of forcing already served networks to migrate to a new regional registry.

That history explains the document's strengths. It asks a candidate to prove broad support from LIRs and ISPs in the proposed region. It requires open and transparent policy development. It calls for neutral and impartial treatment of organizations receiving service. It expects technical expertise, production-grade connectivity, reverse-DNS support, operational staff, funding support, record keeping in English for review by other registries, and confidentiality of registration information.

These requirements were appropriate for deciding whether a proposed registry deserved to receive a region from existing service arrangements. They were not written as a mature enforcement code for a recognized registry decades later.

The text does not define when ICANN may suspend recognition, whether a suspension can be partial, what service elements can be held stable, who may trigger review, how member evidence is gathered, whether domestic courts have priority on corporate questions, what independent reviewer decides factual disputes, what standard of proof applies, how long a cure period should last, or how affected resource holders are protected during review.

The absence is not surprising. In 2001, the problem of a recognized RIR losing effective governance was less immediate than the problem of completing regionalization. But the absence becomes consequential when recognition language is deployed during a mature crisis. AFRINIC was finally recognized in 2005. The ICANN Board resolution of 8 April 2005 recited that AFRINIC had submitted an application and transition plan, that the NRO had recommended final approval, and that ICANN's president had determined full conformance with ICP-2. The Board then proclaimed AFRINIC the recognized RIR for the Africa service region.

That resolution shows recognition at its cleanest moment: an application, a transition plan, favorable peer assessment, IANA review, Board approval and regional service handover. It does not show what happens if, later, governance breaks, an election fails, courts appoint a receiver, members allege improper influence, or critical records must be preserved while the legal entity remains under pressure. A resolution of entry cannot be stretched into a complete exit constitution.

The missing middle is also visible in the Address Supporting Organization arrangements. The ASO gives the numbering community a route into ICANN's structure, and ICANN's Bylaws include the Address Supporting Organization within ICANN's internal architecture. But that does not answer the hard remedial question. A supporting organization can advise and coordinate. It does not automatically create a power to place a registry under supervision, replace management, suspend service rights, compel a transfer of records, or bind a domestic court.

The governance problem is therefore structural. ICANN is not an ordinary external regulator of the RIRs. The RIRs are not mere subcontractors of ICANN. Yet ICANN is not irrelevant, because recognition and IANA-numbering coordination depend on a shared system. The power sits between contract, coordination and public reliance. That is precisely why it needs written limits.

Without limits, recognition language can be overread in two directions. One side can claim that because ICANN recognized a registry, ICANN may command any remedial step necessary to preserve global coordination. The other side can claim that because ICANN lacks domestic corporate power, it must remain silent even when registry failure threatens the numbering system. Both claims are too broad. A recognition rule should occupy the middle: enough authority to verify and preserve the service function, not enough authority to decide every governance or commercial dispute.

The draft replacement work now being discussed by the NRO shows that the community understands this gap. The NRO's draft RIR Governance Document Version 2 explicitly covers recognition, operation and derecognition, not merely entry. It defines derecognition, emergency continuity, emergency operators, RIR services, recognition review, audit, operating requirements and handoff effects. The very existence of that draft is evidence that ICP-2's original entry criteria are no longer adequate as the full rulebook.

AFRINIC made the gap visible

AFRINIC is not the only reason to define recognition power, but it is the clearest stress case. The crisis combined court proceedings in Mauritius, receivership, years without ordinary board governance, election disputes, claims about member voting authority, questions about powers of attorney, allegations of improper influence, litigation involving a major resource holder and repeated interventions by ICANN and the NRO. It showed how quickly a registry crisis can move from domestic corporate law to global service confidence.

The NRO's 14 September 2023 statement on the appointment of an official receiver welcomed court developments that it said could restore AFRINIC to functional governance, allow it to resume full participation in the NRO, and help members keep receiving registry services. The statement also recognized that the receiver was tasked with overseeing elections and that continuity of services mattered during the receivership. That was a continuity posture rather than an attempt to replace AFRINIC.

By June 2025, the issue had become sharper. In the ICANN letter of 25 June 2025 to appointed receiver Gowtamsingh Dabee, ICANN reminded AFRINIC that it had been recognized in 2005 as the RIR for Africa and the Indian Ocean, said AFRINIC continued to have responsibilities under ICP-2, and noted that a compliance review had not yet been initiated. The same letter placed AFRINIC on notice that a review might be necessary because of allegations surrounding the Board election.

The letter was specific in some respects. It identified concerns about voting registration, powers of attorney, access to membership lists, use of an AFRINIC logo in election communications, record keeping and the status of registration data. It tied those concerns to ICP-2 obligations involving community support, equal treatment, impartial treatment, independent open membership and records essential to responsible neutral operations. It required a substantiated response by 26 June 2025 and asked a series of numbered questions.

Yet the letter also illustrates the undefined power problem. It said a compliance review might be necessary, but did not itself define the review's full legal consequence. It reserved rights, opposed proceeding with the election as matters then stood, demanded preservation of records and asked AFRINIC to maintain appropriate customer records for allocations. That is strong pressure.

The public record still left open what formal review would decide, who would review disputed evidence, what interim limits could be imposed, whether a failed review could lead to suspension, whether resource holders could be heard, and what service continuity would look like.

ICANN's 3 July 2025 letter sharpened the point. ICANN said the receiver's response had not supplied sufficient documentation, that annulment of the election did not answer many questions, and that ICANN reserved all rights to initiate a compliance review of AFRINIC regarding areas of potential material non-compliance with ICP-2. It also said the NRO confirmed ICANN was acting in alignment with ICP-2. The letter reinforced the notice function, but still did not provide a settled replacement or suspension rule.

This is the risk. A notice can be necessary and legitimate, especially when the record suggests possible fraud, unfair voting, compromised records or weak neutrality. But a notice that gestures toward recognition status without defined next steps can become leverage in the hands of every actor around the crisis. The incumbent can portray it as overreach. Opponents can portray it as proof that the registry is no longer legitimate. Courts can receive it as a technical warning without knowing its remedial boundaries. Members can be left uncertain whether their services are stable.

AFRINIC shows that recognition power has to handle messy facts. It must interact with domestic courts without being captured by domestic corporate disputes. It must protect ordinary resource holders who are not parties to litigation. It must address election integrity without letting one faction turn global recognition into a campaign tool. It must preserve registry services without freezing reform. It must require records without exposing confidential member information. It must support continuity without converting ICANN into a substitute board.

The current record shows ICANN trying to navigate those tensions, but trying under an incomplete rule. Good intentions are not enough. The next crisis should not depend on how precisely one letter is drafted under time pressure.

Notice is not a trigger

A defined recognition power begins by separating notice from trigger. Notice tells an institution that concerns have arisen and that records must be preserved. A trigger is the factual threshold that opens a formal review with consequences. If these are blurred, ICANN can appear to initiate discipline before a review exists, while the registry can dismiss serious evidence as mere correspondence.

The notice stage should be broad and fast. ICANN should be able to send a preservation notice when credible information suggests a material threat to RIR services, neutrality, records, member rights, technical continuity or compliance with recognized obligations. The notice should identify the concern, affected duties, requested records, confidentiality treatment and response time. It should state clearly whether formal review has begun. If not, it should say so.

The trigger stage should be narrower. A formal review should require an articulated basis, such as failure to maintain effective governance, inability to provide critical registry services, material failure to preserve registration records, credible evidence of substantial improper influence over registry service decisions, sustained inability to treat members impartially, failure to maintain sufficient financial or operational independence, refusal to cooperate with a properly initiated audit, or a court-confirmed condition that prevents core services from continuing.

The trigger should also name the service connection. Not every governance defect should activate recognition consequences. A late annual report, a disputed meeting procedure, a controversial policy proposal or a single lawsuit may matter, but it should not by itself threaten recognition. The question is whether the defect endangers the registry's ability to provide stable, reliable, secure, accurate and accountable registry services, or whether it undermines the legitimacy of governance in a way that affects those services.

This service connection prevents mandate laundering. ICANN should not use recognition to decide who is right in an ordinary commercial dispute, to police every local election rule, to approve one policy faction over another, or to enforce a preferred regional political settlement. Its legitimate interest is the continuity and neutrality of the number-resource system. The more distant a concern is from that function, the stronger the justification must be before recognition language is used.

The trigger should also identify the evidence source. A complaint from a single member, an article, a court filing, a letter from another RIR, an independent audit, a member petition and observed service failure are not equal. Each can start inquiry, but a formal review needs a written evidence statement. That statement should distinguish verified facts, allegations requiring testing, external legal orders, technical observations and matters outside ICANN's competence.

AFRINIC's 2025 election concerns show why this matters. Alleged forged powers of attorney, unusual voting authority, possible access to member lists and unclear record backups are all serious. Some are election integrity questions. Some are member-rights questions. Some are registry-record questions. Some are possible criminal or domestic legal questions. A precise trigger statement would say which questions ICANN can evaluate directly and which questions it treats as risk indicators while waiting for courts or investigators.

The rule should also define when escalation pauses. If a registry supplies requested records, announces a credible correction, permits an independent election audit, preserves data, maintains services and agrees to a verifiable cure plan, the review should move into remediation rather than automatically intensify. The goal should be compliance and continuity, not punishment for its own sake.

Conversely, failure to answer, destruction of records, false statements, service interruption or ongoing manipulation should move the review forward. That should happen because the published trigger and response standard say so, not because one side has louder public rhetoric.

A suspension power must name the service it suspends

The most dangerous undefined tool is suspension. Withdrawal is dramatic and rare; ordinary notice is reversible. Suspension sits in the middle and can do heavy damage if it is vague. Does it suspend voting recognition? New allocations? Transfers? reverse-DNS changes? RPKI issuance? NRO membership rights? IANA allocations? public trust in the registry's database? The word cannot be used safely unless it names the service.

A legitimate interim power should be modular. If the defect concerns a board election, the interim step might be limited to preventing certification of a contested result until member authority is verified. If the defect concerns registration data integrity, the interim step might require preservation, backup and independent reconciliation while routine non-disputed updates continue. If the defect concerns payments or corporate paralysis, the interim step might authorize a narrow continuity budget. If the defect concerns RPKI or reverse DNS, technical continuity should be preserved unless there is a direct security reason to freeze a change.

Resource holders should not pay for institutional ambiguity through unnecessary service freezes. A holder that needs a legitimate contact update, reverse-DNS repair or RPKI renewal should not be stranded because an unrelated election dispute is under review. Transfer requests may require caution if authority files are in doubt, but even then the rule should distinguish new contested transfers from ordinary corrections and previously approved changes.

The rule should also protect non-parties. In a regional registry crisis, most affected networks are not litigants. They need stable records, functioning help desks, billing clarity, abuse contactability, route-origin certification, reverse-DNS service and the ability to prove their address status to suppliers and customers. Interim recognition measures should be designed around the least disruptive service control that addresses the identified risk.

This is where the draft NRO RIR Governance Document points in the right direction. Its operating requirements define RIR services broadly enough to include delegation, registration, directory and related technical services. It also defines emergency continuity, emergency operator and handoff concepts. Those definitions should be refined and adopted in a way that makes interim action service-specific rather than rhetorical.

Suspension also requires publication discipline. A public notice should not merely say that a registry is under review. It should identify affected and unaffected services. It should say whether members can still pay fees, vote, request updates, submit transfers, use RPKI, change reverse DNS, rely on existing records and contact staff. If confidentiality limits detail, the notice should say what cannot be disclosed and why.

Without service-specific suspension, recognition warnings create avoidable market fear. Buyers ask whether any transfer from the region is unsafe. Cloud providers ask whether bring-your-own-address entries should be delayed. Lenders ask whether records might later be invalid. Courts ask whether they are preserving assets or damaging public infrastructure. A narrow notice reduces this uncertainty.

It also prevents overreach. ICANN should not obtain, by implication, a broad power to control regional registry business because one service is at risk. If only election certification is unsafe, only election certification should be constrained. If only records need preservation, only records should be preserved and audited. A modular suspension power makes proportionality real.

Withdrawal must begin with evidence, not exhaustion

Withdrawal of recognition should be a last resort, but "last resort" is not a standard unless it is tied to evidence. Institutions often say that severe action is reserved for extreme cases. When the case arrives, the same institutions disagree about what counts as extreme. A credible rule must say what must be proven before withdrawal can be considered.

At minimum, withdrawal should require a written proposal identifying the specific obligations failed, the evidence supporting each failure, the service consequences, the attempted cure steps, the reasons lesser measures are limited public evidence and the continuity plan that would preserve critical services. A registry facing withdrawal should receive the proposal, have a defined response time, and be able to submit evidence. Other RIRs should review independently and publish reasons for their recommendations.

That structure appears in the NRO draft. Version 2 provides that a proposal for derecognition may be submitted by any RIR or group of RIRs, a large group of members of the affected RIR, or ICANN; the proposal must identify reasons and specific failed provisions; the affected RIR must receive a reasonable response period; the other RIRs must independently consider and publish recommendations; unanimity among the other RIRs is required before referral to ICANN; and ICANN must publish its decision and reasons. It also says ICANN has no power to recognize or derecognize an RIR unless it has received an approved proposal.

Those are important limits. They turn recognition power from unilateral pressure into a staged, reasoned decision. They also reduce the risk that ICANN alone becomes the political center of a regional crisis. If the other registries must publish reasons, the decision becomes a system judgment rather than a single-office letter.

But evidence must still be practical. The rule should not require public release of every member record, legal opinion or confidential security detail. It should require a public evidence map. That map can identify categories, dates, decision points, omitted records, disputed facts, confidential exhibits and the reason each item matters. Redaction should protect privacy and security without hiding the causal chain.

Withdrawal should also distinguish institutional failure from contested rights. A member dispute over a particular allocation, a commercial disagreement about IPv4 leasing, a court fight over board authority or a disputed election can be evidence of institutional stress. It is not automatically proof that the registry can no longer provide RIR services. The withdrawal standard should require that the failure be material to the registry's continuing service role.

This threshold protects every side. It protects ICANN from accusations that recognition is a factional weapon. It protects the affected registry from opportunistic petitions. It protects members from indefinite tolerance of an institution that cannot function. It protects courts from being asked to infer global technical consequences from local corporate pleadings. It protects the market from rumor-driven discounts.

The evidentiary rule should include independent review where facts are contested. If the allegation concerns technical service failure, an independent operational auditor may be appropriate. If it concerns governance manipulation, an election or corporate-governance reviewer may be needed. If it concerns records, a data custody and reconciliation auditor may be needed. One reviewer cannot resolve every kind of fact. The rule should match expertise to claim.

Remedy matters because recognition is leverage

Procedure is incomplete without remedy. If ICANN or the RIRs can review and make findings, but the affected registry or its members cannot obtain a meaningful correction when the review is wrong, recognition remains leverage rather than accountable authority. Remedy does not mean that every dissatisfied party can block action forever. It means that high-consequence decisions have routes for correction, stay and review.

For notice and information requests, remedy may be modest. The registry should be able to ask for clarification, confidentiality limits, reasonable extensions and a written statement of how requested information relates to recognized duties. If ICANN overreaches, the record should show the objection and response.

For formal review, remedy should be stronger. The affected registry should receive the evidence statement, have a chance to answer, know the identity and terms of any independent reviewer, challenge conflicts, submit counter-evidence, and receive a reasoned finding. Members and resource holders should have a route to submit evidence when their services or governance rights are directly affected, without turning the review into an unmanageable open forum.

For interim measures, remedy should include proportionality review. If a service freeze is imposed, the registry or affected holders should be able to ask whether a narrower safeguard would address the risk. If a transfer is paused, the applicant should know whether the pause is due to authority uncertainty, record preservation, dispute notation or another stated reason. If an emergency service arrangement is activated, the affected community should know its scope and expected duration.

For withdrawal, remedy must include independent and time-sensitive review. ICANN's existing accountability routes may provide some review, but number-resource recognition has distinctive consequences. A general corporate accountability route may not be enough if the issue is whether millions of resource records, technical services and regional member rights remain portable during transition. The rule should define reviewability as part of recognition power, not as an afterthought.

Remedy also requires cure. A non-compliant registry should receive a realistic path back unless immediate continuity demands a temporary substitute. The NRO draft's rehabilitation clause is important here. It creates a presumption in favor of helping a non-compliant RIR cure and says derecognition should be a last resort where the harm of tolerating non-compliance outweighs the benefits of continuing. That is the right principle. It must be paired with measurable cure conditions: governance restoration, record backup, independent audit, service metrics, financial controls, conflict rules, member notice and periodic reporting.

Without cure conditions, "support" becomes vague reassurance. With them, recognition power becomes disciplined. The affected registry knows what must be done. Members know what progress looks like. ICANN and the other RIRs can justify either de-escalation or escalation. Courts can see the difference between institutional recovery and continuing failure.

The key is that remedy should not wait until final withdrawal. By then, too much value has moved. An incorrect warning can already raise transaction costs. An overbroad interim freeze can already harm networks. A poorly explained review can already polarize the region. Remedy must be built into every stage because recognition leverage begins before final decision.

The boundary between coordination and political pressure

The hardest line is between legitimate coordination and political pressure. ICANN has a real interest in the stability of the number-resource system. The RIRs have real regional autonomy. Local courts have real authority over incorporated entities and disputes before them. Members have real rights under bylaws, service agreements and applicable law. Resource holders have reliance interests in stable records. None of these interests cancels the others.

Recognition power should therefore be framed as functional, not political. ICANN should be able to ask whether the recognized registry can provide RIR services neutrally, accurately, securely and continuously. It should be cautious about deciding who should win an election, which faction best represents a region, whether a commercial IPv4 business model is good policy, or how domestic membership law should be reformed beyond the service implications.

The 2025 AFRINIC letters show the tension. ICANN's concerns about powers of attorney, member lists and record backups were tied to core duties. Its concern that AFRINIC's operations remain independent of influences that would move policy counter to global numbering coordination was more sensitive. Influence over policy can be a real risk if one actor captures a registry. But broad language about undesirable policy direction can sound like ICANN policing regional policy outcomes rather than protecting neutral service.

A defined rule can solve this by requiring a service-based harm statement. If influence is alleged, the question should be whether it compromises impartial treatment, member rights, record integrity, policy-development openness, financial independence or continuity. ICANN should not claim power simply because it dislikes a policy position. It should claim power only when the method of control or the resulting service risk violates a recognized duty.

The same boundary applies to courts. A court may appoint a receiver, restrain a corporate act, order preservation, address fraud allegations or interpret membership rights. ICANN should respect those orders while explaining technical continuity. It should not ask a court to treat recognition as a trump card over domestic law. It should, however, make clear when a proposed corporate remedy would impair RIR services, destroy records, confuse resource holders or create region-wide instability.

Political pressure also comes from the opposite direction. A troubled registry can invoke stability to resist every accountability demand. It can say that any review endangers the internet, that any challenge to its board undermines community self-governance, or that every complainant is hostile to the region. A defined recognition rule prevents this defensive misuse as well. Stability is protected by evidence, service continuity and cure, not by institutional immunity.

The principle is simple: recognition protects the public function, not the incumbent office. It protects the ledger, services, members and continuity of number-resource management. It does not protect a board, receiver, faction, litigant, commercial strategy or political narrative from scrutiny. That principle should be written into the rule.

What a defined recognition power would contain

A credible recognition rule would have ten parts. First, it would define recognized duties in service terms: registration, allocation, directory services, reverse DNS, RPKI or certification services where applicable, member governance, record keeping, policy development, confidentiality, financial independence, operational independence and continuity.

Second, it would define triggers. A trigger should be a credible, documented risk to one or more recognized duties, not merely reputational discomfort. The rule should include service failure, loss of effective governance, record-integrity risk, material refusal to cooperate with audit, substantial improper influence, inability to treat members impartially, loss of operational independence, court-ordered paralysis of core services, and serious failure of continuity controls.

Third, it would define the notice stage. The notice should identify the concern, the linked duty, the evidence currently known, the requested response, the preservation duties, confidentiality handling and whether formal review has begun. It should be published unless confidentiality requires a temporary redaction.

Fourth, it would define formal review. Review should start with a written statement, an evidence map, a response period, reviewer independence, conflict rules, publication of terms, and expert matching to the dispute. A governance allegation, technical service allegation and record-custody allegation require different expertise.

Fifth, it would define interim safeguards. Interim measures should be service-specific, time-limited, reviewable and least disruptive. They should preserve routine services for non-disputed resource holders wherever safe.

Sixth, it would define cure. A cure plan should have measurable conditions, dates, reporting duties and verification. It should distinguish immediate actions such as preserving backups from deeper actions such as restoring elected governance or updating controls.

Seventh, it would define emergency service support. If an emergency operator is needed, the rule should state scope, authority, data access, privacy, duration, community notice, funding, limitation, review and handback. An emergency operator should not become a permanent replacement by inertia.

Eighth, it would define withdrawal. Withdrawal should require an identified proposal, specific failed duties, published reasons, other RIR recommendations, ICANN decision reasons, continuity plan and independent review. It should not be available merely because the affected registry is unpopular or politically inconvenient.

Ninth, it would define resource-holder protections. Existing records, pending transfers, RPKI material, reverse-DNS delegations, disputes, fees, contact updates and confidential data must have continuity rules. Holders need to know how their rights and services travel during review or transition.

Tenth, it would define accountability for ICANN itself. ICANN's recognition decisions should be reasoned, reviewable, proportionate and constrained by mission and agreement. If ICANN exceeds scope, delays without reason, ignores evidence or applies inconsistent standards, the affected parties need an effective challenge.

The NRO draft already contains several of these elements, especially derecognition proposal requirements, audit, emergency continuity, operational requirements, rehabilitation, handoff and readiness. Its next test is precision. The rule should not only declare that emergency continuity may occur; it should define the actual migration file, service inventory, record custody, implementation state, technical trust state, communications plan and review clock.

The test for ICANN now

ICANN's legitimacy in the RIR layer will depend less on whether it can write a forceful letter and more on whether it can bind itself to a predictable rule. The temptation in crisis is to preserve discretion. Discretion feels useful when facts are incomplete and speed matters. But undefined discretion is exactly what converts recognition into leverage. The stronger long-term move is to accept limits before the next case.

The AFRINIC letters show that ICANN can identify concrete concerns and demand evidence. That is necessary. They also show why the rule cannot stop at warnings. A warning about possible review affects markets and courts before final review begins. If ICANN wants that warning to be seen as public-interest coordination rather than institutional pressure, it should publish the exact route from notice to review, cure, emergency support and possible withdrawal.

The other RIRs should want the same. A clear rule protects them from political contagion. If one registry fails, the others should not be forced into improvisation or exposed to accusations that they are rescuing an incumbent club. They should be able to point to predefined duties, an evidence standard, an emergency support role and a handoff rule. Their legitimacy depends on showing that peer coordination is not peer protection.

Members and resource holders should also want clarity. A registry crisis should not make every address record in a region more expensive to rely on. Clear triggers and service-specific safeguards reduce private risk premiums. They let lawyers, engineers, financiers and public agencies ask better questions. Is my service affected? Is my record preserved? Is my transfer paused? Who has authority to sign? What evidence is needed? What review exists? Those questions are practical, not theoretical.

The answer should be available before a crisis letter arrives. Recognition is too important to remain a residual power inferred from old entry criteria. It should be a complete public discipline: entry, operation, review, cure, emergency continuity, withdrawal, handoff and remedy.

If the rule is written narrowly, it will not weaken ICANN. It will strengthen the part of ICANN that matters here: the ability to coordinate unique identifiers without becoming a political supervisor. Clear recognition power is not a centralization project. It is a restraint project. It tells every registry that recognition is real, every member that continuity is protected, every court that technical reliance is visible, and every claimant that crisis cannot be used to seize the number-resource system by rhetoric.

The undefined power worked only while no one tested it. AFRINIC tested it. The next rule should make sure that recognition remains a safeguard, not a threat whose meaning is discovered only after damage has begun.

Sources