Summary

  • AFRINIC’s proposal register dates AFPUB-2020-GEN-006-DRAFT03 to 22 November 2021; a later interlocutory judgment records that AFRINIC withdrew a court-related undertaking on 25 November amid a dispute over membership, account status and delegated-resource administration.
  • The three-day proximity proves neither causation nor motive. Its significance is institutional: a formal route for moving registry administration was advancing inside the same private registry that was then asserting contested control over whether a member and its resources could remain administratively active.
  • A transfer policy is necessary because private parties cannot safely rewrite authoritative registries by contract alone. Yet the registry’s proper role is narrow—authentication, consistency, atomic handoff, correction and continuity—not judgment over business merit, price, customer location or regional loyalty.
  • The proposal’s later ratification on 4 February 2026 confirms that the 2021 draft completed one formal stage. It does not establish implementation, reciprocal recognition by another registry, a completed transfer or a measurable economic result.
  • A credible exit route would use a reason-coded, time-bounded and independently reviewable receipt that preserves the last verified operational state during a genuine dispute and prevents the incumbent registry from becoming the final judge of departure.

The three days that frame the question

AFRINIC’s proposal register assigns 22 November 2021 to the third draft of the Number Resource Transfer Policy. The document’s exact identifier is AFPUB-2020-GEN-006-DRAFT03. Three days later, on 25 November, AFRINIC withdrew an undertaking it had given in litigation after a July court order restrained action connected with membership termination, an account freeze and resource reclamation. The later interlocutory judgment records both events in the litigation sequence. The proposal register and the judicial record therefore place two different institutional tracks almost side by side: one concerned a formal route for transferring registry administration; the other concerned AFRINIC’s attempted control over the continuing administrative position of a member and delegated resources.

That proximity must be handled with discipline. It does not show that the policy proposal targeted Cloud Innovation. It does not show that the withdrawal answered the draft, that the draft answered the litigation, or that either event caused the other. It establishes no motive, coordination or bad faith. A calendar can identify an institutional setting without proving a private intention.

The relevant point is narrower and more consequential: at the moment a transfer route entered a formal approval path, AFRINIC was also a party to a dispute in which its ability to terminate membership, freeze an account and reclaim delegated number resources was being contested.

This is not a story about a suspicious date. It is a test of institutional design. Transfer rules matter most not when relations between a registry and a holder are calm, but when trust has failed, the registry is conflicted or continuity is threatened. If a route works only while the incumbent approves of the holder’s conduct and business, it is not fully an exit route. It is continued dependence presented as mobility. Conversely, if the registry is stripped of every role, a private agreement can leave two authoritative ledgers inconsistent, permit duplicate claims or fracture the operational services attached to a prefix or ASN.

The hard problem is to preserve coordination without converting coordination into dominion.

The November sequence supplies a real setting in which to ask that question. A proposed transfer path and a contested resource-control path occupied the same institution at almost the same time. The case does not resolve the meaning of the transfer text. The transfer text does not resolve the case. Taken together, however, they expose the design problem that an abstract policy debate can conceal: who controls exit when the recordkeeper is itself one of the disputing parties?

Correcting the proposal before interpreting it

The identifier correction is foundational. The subject is AFPUB-2020-GEN-006-DRAFT03, dated 22 November 2021 and recorded as later ratified. It is not AFPUB-2019-V4-003-DRAFT03, a separate proposal from September 2020. Similar labels can invite a false continuity, especially when several transfer drafts moved through AFRINIC’s records across adjacent years. But the two lineages cannot be blended. Dates, clauses, procedural records and conclusions belonging to the 2019-V4-003 sequence do not migrate into the 2020-GEN-006 sequence merely because both concern transfers and both reached a third draft.

The correction changes more than a citation. It fixes the institutional act under examination. This article concerns the entry, on 22 November 2021, of a particular Draft 3 into the formal route that eventually reached Board ratification. It does not assess the redlines of the other Draft 3, the compatibility analysis associated with earlier drafts, the subsequent Draft 4 in that other lineage, or an appeal concerning it. Those are different records with different procedural questions. Importing them would create an apparently rich narrative by assembling facts that do not belong to the same instrument.

Precision is especially important because a transfer policy accumulates meaning through version, date and status. A proposal may be published without having secured consensus. Consensus may be claimed or assessed without a Board having ratified the text. Ratification need not mean the rule has taken effect. Effective force need not mean operating procedures, counterpart agreements and technical systems are ready. A working route to a specific destination requires still more: reciprocal compatibility, coordinated execution and an actual request that meets the applicable conditions.

When identifiers are confused, those distinct states collapse into a single misleading impression of “the policy.”

The current AFRINIC record says the corrected Draft 3 was ratified by the Board on 4 February 2026. That fact gives the 22 November 2021 text a later lifecycle state. It shows that the draft eventually completed a formal approval route. It does not establish the date of implementation, the readiness of any counterpart registry, an operating manual, a service commitment, or a completed transfer. Nor does it invite a detour into the Board act itself. The proper use of the 2026 date here is as a boundary marker: proposal in 2021, Board ratification in 2026, every later operational state still requiring its own evidence.

The dispute supplies context, not a verdict

The litigation record must be kept within similarly exact limits. AFRINIC’s public Board record says Resolution 202107.623 addressed the Cloud Innovation matter or cases of a similar nature through four linked steps: terminate the concerned Resource Member’s membership, freeze its account, notify users of imminent resource reclamation, and reclaim the number resources after 90 days. A Mauritius court injunction on 13 July restrained action related to that resolution. AFRINIC gave an undertaking on 15 July, and the initial interim order was then discharged.

The later interlocutory judgment records withdrawal of the undertaking on 25 November and a later interim restraint while the substantive dispute remained for final determination.

Those facts show the range of control AFRINIC was attempting to exercise. Membership, the account, registry access and the continued administrative treatment of delegated resources were connected in one chain. That is enough to explain why exit was not a theoretical concern. A holder disputing the registry’s actions could face a situation in which the institution maintaining the record was also seeking changes with potentially serious operational consequences. The holder’s need for an alternative administrator would then be strongest at exactly the moment the incumbent’s consent might be least neutral.

But the interlocutory record does not decide whether Cloud Innovation breached its Registration Service Agreement. It does not decide whether AFRINIC ultimately possessed a contractual entitlement to terminate membership or reclaim resources. It does not decide ownership of the resources. An interim order and an undertaking are procedural events, not a final adjudication of the underlying rights. It would therefore be wrong to turn this analysis into either a victory narrative for the member or an accusation of unlawful conduct by the registry.

The bounded conclusion is institutional rather than merits-based. AFRINIC was a private party asserting one view of its contractual and administrative powers while another private party contested that view before an ordinary court. At the same time, AFRINIC was the registry through which the proposed transfer route would have to be recognised and recorded. No finding of misconduct is needed to see the conflict-of-role risk. Sound institutional design assumes that honest organisations can become interested parties. It constrains authority before anyone has to prove malice.

That distinction also prevents a common analytical shortcut. The argument for portability does not depend on proving that the holder was right in the dispute. Even a holder that may ultimately lose on some contractual issue is entitled to a process that separates disputed private rights from unilateral alterations to shared technical records. Equally, preserving the last verified operational state during a dispute does not decide the merits for the holder. It simply avoids allowing control of the ledger to become a substitute for adjudication.

What the transfer architecture offered—and withheld

AFRINIC’s later account of the ratified architecture describes more than one route. It covers intra-regional transfers, transfers from legacy holders to members, and reciprocal inter-RIR transfers for IPv4 number resources and Autonomous System Numbers. That breadth matters because portability is not one transaction. A move between two parties under one registry differs from a move across two registries. An ASN is not an IPv4 block. A legacy resource carries a different administrative history from a resource issued by AFRINIC. Any honest assessment must preserve those distinctions.

The architecture did not create an unconditional ability to leave. It classified resources and treated categories differently. In particular, AFRINIC-issued IPv4 resources were restricted from leaving the service region, while outbound movement was available only to specified eligible categories, including some legacy or externally received resources. The text made AFRINIC’s written approval and registry update part of the recognised route, declined to recognise transfers conducted outside the approved process, and subjected incoming resources to the AFRINIC policy surface. Inter-RIR movement also depended on reciprocal policy compatibility.

Several of those elements can serve legitimate recordkeeping ends. The registry must know whether the source has recognised control of the exact resource. It must distinguish an active allocation from a reserved range, a legacy record or a block previously received from elsewhere. A destination registry must be able to accept the resource and recipient under its own applicable rules. Both registries must avoid a period in which each ledger, or neither ledger, presents itself as authoritative. Written approval can provide a timestamped execution record. Non-recognition of an improvised transfer can protect users from contradictory claims.

Yet the same elements can also preserve institutional leverage. A regional restriction may prevent a holder from choosing an otherwise capable administrator. A broad approval requirement may allow the incumbent to assess matters beyond identity, control and technical eligibility. Attaching all incoming resources to the incumbent’s policy surface can convert entry into durable dependence. A reciprocity requirement can identify a real coordination problem, but it can also become an opaque assertion unless the exact counterpart, policy version, direction and technical readiness are visible.

The critical distinction is therefore not “approval or no approval.” It is the substance and legal character of approval. A ministerial approval records that specified, objective conditions have been met and triggers a defined handoff. A discretionary approval asks whether the institution wishes to permit the transaction. The first can be reason-coded, time-bounded and reviewed against a stable record. The second leaves the holder’s exit exposed to judgments that may have little to do with registry accuracy.

This distinction is particularly important for outbound transfer. If the route is meant to discipline the incumbent relationship, the incumbent cannot possess an unrestricted veto over departure. Otherwise the right exists only for holders that do not need it. Genuine portability requires that a qualified holder be able to change registry administration without surrendering operational identity and without inviting the old administrator to decide whether the holder’s business model, customers or regional commitments deserve approval.

Number resources as operational identity

The underlying principle is that number resources are operational assets and network identities. A prefix is woven into routing, access lists, customer configurations, reputation systems, security controls and business continuity. An ASN can be embedded in routing policy and external relationships. Moving away from a registry should not require a network to discard that identity and renumber merely because the incumbent service relationship has failed.

Portability is therefore a structural right of exit. It is not a claim that registry data can be altered casually, nor a promise that every category of resource may move everywhere without conditions. It is the proposition that registry administration should follow recognised operational and legal reality instead of trapping a holder inside an institutional relationship. The possibility of exit disciplines a service provider even when no transfer occurs. A registry that knows holders can leave has a stronger reason to provide accurate, predictable and impartial service.

This view also fixes AFRINIC’s institutional character. AFRINIC is a private technical bookkeeper, service provider and coordinator. It is not the sovereign of African address space. It does not legislate in the public-law sense, police lawful business models, prosecute policy offences, punish unpopular use, confiscate assets or sit as a public-law court. Describing these powers as absent is not rhetorical excess; it prevents a private recordkeeping function from silently acquiring the attributes of government.

The region is a service area, not title. The fact that a registry serves Africa does not make number resources the political property of a region or of a self-declared community. A registry entry describes a recognised state: which holder controls a resource, which contacts are current, which administrative authority maintains the record, and what dependent services are associated with it. The entry does not manufacture ownership by the act of recording. The ledger follows reality; it does not create a sovereign entitlement to command that reality.

None of this makes the registry irrelevant. The authoritative record is valuable precisely because networks need a coherent view of unique identifiers. But uniqueness coordination must stay thin. AFRINIC may verify recognised control, keep contacts accurate, isolate a documented dispute, coordinate a transfer, retain history, correct errors and protect continuity. It may not use the indispensability of the record as proof that it owns the thing recorded or may adjudicate every controversy touching it.

The distinction resembles the difference between maintaining an accurate register and deciding the commercial wisdom of the underlying transaction. The first is necessary infrastructure. The second belongs to private parties operating under ordinary law, with courts available for genuine disputes. A registry can refuse to enter contradictory or technically impossible states. It should not refuse because it dislikes the price, the financing, the customer base, the use of leasing, or the holder’s decision to seek service elsewhere.

Why a formal route is still necessary

The strongest case for a formal transfer policy begins with the risks of an informal market. A private sale agreement cannot, by itself, update two authoritative registries. It cannot guarantee that a source claiming to transfer a block is the recognised holder. It cannot prevent the same resource from being promised twice. It cannot ensure that the destination registry has authenticated the recipient or that the two systems agree on the instant when administrative custody changes.

The handoff also reaches beyond a row in a database. WHOIS or RDAP contact information must change coherently. Reverse-DNS delegation may need coordinated treatment. RPKI state must not be left in a contradictory or unusable position. Transfer history should show what changed without exposing the record to silent revision. If one technical step fails, the parties need a defined rollback or a safe intermediate state. If an error is discovered later, the record needs a correction trail rather than an unexplained overwrite.

Fraud resistance and dispute isolation are legitimate as well. A documented dispute over control cannot be ignored simply because one party has presented a transfer agreement. The registry should be able to mark the dispute, preserve evidence and prevent a race to create irreversible registry facts. The destination needs assurance that it is not accepting a resource subject to an unresolved duplicate claim. The source needs assurance that the request came from the recognised holder. These are not excuses for arbitrary power; they are the ordinary safeguards of reliable coordination.

Inter-RIR transfers add another layer. Ratification by AFRINIC alone cannot make a route executable with every possible counterpart. The other registry must have a compatible policy, accept the relevant resource class and direction, recognise the recipient, and coordinate the operational cutover. Compatibility must be specific. The existence of two transfer policies in the abstract does not establish that their categories, eligibility tests, documentation and timing align for a particular transaction.

A written, predictable route is thus better than covert transfers, contradictory private records or unpublished staff exceptions. It can replace personal favour with objective evidence. It can give both sides a durable receipt. It can expose incompatible policy versions before a transaction reaches the point of operational risk. It can reduce the chance that the public record diverges from the network reality it is meant to describe.

The benign case deserves to be stated fully because an argument for exit that ignores these tasks will fail in practice. Portability cannot mean that anyone may demand an immediate ledger change without proving control. Independence from institutional discretion does not mean independence from evidence. The correct response to an overbroad gate is not to abolish the gatehouse; it is to specify the few questions the gatekeeper may ask, require timely answers, record the reasons and provide neutral review.

The line between verification and veto

The legitimate questions are concrete. Is the source the recognised holder of the exact prefix or ASN? Is the resource within a category eligible for this direction of transfer under the stated policy versions? Is there a documented adverse claim, and if so, what neutral interim state preserves continuity? Has the destination authenticated the recipient? Do the source and destination policies match for this resource class and direction? Can the registry and its dependent services change atomically, with a tested rollback?

Questions about business merit fall outside that role. The registry should not decide whether the agreed price is attractive, whether financing is prudent, whether leasing is desirable, whether customers are located in a preferred place, whether a holder has shown sufficient regional loyalty, or whether the institution approves of the business model. Those matters neither prove recognised control nor protect the uniqueness of the registry. When they become grounds for refusing to record a transfer, bookkeeping becomes a form of private economic permission.

Regional restrictions require this same scrutiny. A service region can define which registry normally administers a holder or how regional coordination is organised. It does not itself create ownership of the resources associated with that service. If a category cannot leave only because the incumbent declares that resources issued in its region must remain under its administrative surface, the condition preserves lock-in unless it can be tied to a necessary and proportionate registry function. Conservation preferences or regional ideology are not substitutes for title.

Written approval is legitimate only when it confirms objective compliance. The reason should correspond to a published condition. The institution should respond within a defined period. A refusal should identify the exact missing evidence or incompatible state. The holder should have access to an independent reviewer capable of suspending contested changes. Silence, unexplained delay or a general claim of discretion should not be allowed to accomplish what an express denial could not defend.

The conflict becomes acute when AFRINIC itself is a disputing party. An institution may authenticate records in its possession, but it cannot be the final adjudicator of the private controversy in which it participates. If it may terminate membership, freeze an account or seek reclamation and then use those same disputed steps to defeat transfer eligibility, the transfer route cannot reliably serve as an escape from the conflict. The incumbent would be able to create the condition that blocks departure and then cite that condition as neutral registry fact.

The safer rule is to preserve the last verified operational state while the dispute is examined. That does not guarantee transfer. It prevents a unilateral administrative change from prejudging the result. An ordinary court can decide contractual rights. A genuinely independent reviewer can determine whether the registry has applied an objective transfer condition correctly. The reviewer should possess suspensive authority so that review is meaningful before an irreversible registry or dependent-service change occurs.

This approach protects both sides. A holder cannot manufacture portability by ignoring a genuine competing claim. A registry cannot manufacture ineligibility by turning its own contested action into a final fact. The technical record remains stable, the evidence remains visible, and the merits proceed in a forum that does not depend on one party’s control of the ledger.

Exit as an economic discipline, not a promised outcome

The practical stakes follow from that institutional design. A credible exit option can expand the set of potential counterparties and registry relationships available to a holder. It can reduce the fear that a dispute with one service provider will force the network to renumber. It can improve continuity planning because administrative custody is not treated as inseparable from the incumbent. It can also strengthen bargaining discipline: a service provider must retain trust rather than rely entirely on the cost of leaving.

A conditional or opaque route produces the opposite pressures. A bidder may need more diligence to determine whether the incumbent will recognise the transaction. Parties may need additional contractual conditions, escrow arrangements or legal opinions. Timing becomes harder to predict when reciprocal compatibility is asserted at a high level instead of shown through exact policy versions and directions. A regional restriction can shrink the field of eligible recipients. A discretionary approval gate can leave financing or operational plans contingent on an interested institution’s decision.

These are mechanisms of exposure, not evidence that a measured loss occurred. The available record does not establish a completed transfer under the 2021 Draft 3, a rejected transaction, a route change, an outage, a customer count, a price effect or a monetary loss attributable to the proposal. It would be unsound to convert plausible transaction costs into invented figures. The governance concern exists even without a quantified incident because the allocation of decision power affects what holders and counterparties must plan for.

Portability’s discipline is partly counterfactual. A holder may never leave, just as a customer may never invoke a termination clause. The credible ability to do so changes the relationship. If departure is technically possible but institutionally subject to an undefined veto, the disciplining effect weakens. A nominal route can then coexist with practical lock-in.

This is why implementation evidence matters. A ratified text can describe an exit path on paper. Only operating rules, counterpart readiness, transparent reasons, service timing, correction records and completed cases can show how it behaves. The absence of such evidence in this record does not prove failure. It means the conclusion must stop at institutional design and the later fact of ratification.