Summary

  • AFRINIC’s public D3 proposal page says it does not recognise transfers outside approved transfer policies and requires organisations holding the resources to return them to the appropriate registries. The sentence appears in explanatory provision text, not as a separately numbered rule in sections 3.1 to 3.6.
  • A separate staff assessment says transfers made without AFRINIC’s prior approval would be non-compliant and reclaimed, and recommends adding materially that consequence to section 3. The checked material does not prove that reclaim was independently enacted as a numbered rule or that it has been executed.
  • AFRINIC may preserve its last verified record, reject unsupported evidence, prevent duplicate recording and flag fraud or court holds. It is nevertheless a private bookkeeper and coordinator, not a sovereign, regulator, police force or punishment authority. Its record can describe a conflict; it cannot abolish contractual, beneficial, judicial or operational reality by declaration.
  • The missing institutional bridge is a correction forum: notice, a stated evidentiary deficiency, a cure period, conflict metadata, interim continuity, written reasons and genuinely independent review before any irreversible registry action.

There are at least three realities in any disputed number-resource transfer. There is the transaction the parties say they made, supported or undermined by instruments, authority, consideration and applicable law. There is the operational state: who is using the addresses, which contacts can respond to incidents, and which associated systems still point to a previous holder. Then there is the registry state that AFRINIC maintains and publishes through its own services. Good administration tries to bring those realities into accurate alignment. Bad administration assumes that the last of them creates the other two.

The distinction becomes decisive when AFRINIC says it will not recognise a transfer performed outside approved policy. A registry cannot responsibly accept every private assertion. It needs evidence that the source was authorised, that a claimant is who it says it is, that the request is not duplicative or fraudulent, and that a court has not restrained action. It may therefore leave the last verified holder in its record while a claim remains unsupported. That is a legitimate act of bookkeeping. It protects uniqueness and avoids presenting an untested assertion to users of the common record as settled fact.

But non-recognition has two possible meanings. In the narrow meaning, AFRINIC is saying: “We have not verified enough to change our record.” In the expansive meaning, it is saying: “What we do not record does not exist, and we may take the resource away.” The first meaning is compatible with a thin coordinating function. The second converts dependence on a shared registry interface into an instrument of economic force.

The D3 material sits directly on that fault line. Its public proposal page states that AFRINIC does not recognise transfers outside approved transfer policies and requires organisations holding such resources to return them to the appropriate registries. That consequence is the central subject here. It is not a judicial holding that every outside-process agreement is void. It is not proof that AFRINIC owns the numbers. It does not establish that a contractual exchange, beneficial interest, court-recognised claim, operational handoff or routing fact evaporates because AFRINIC has declined to update its database.

The exact location of the language matters. The non-recognition and required-return sentence appears in explanatory provision text on the public proposal page. Separately, the staff assessment says that resources deemed transferred without AFRINIC’s prior approval would be non-compliant and reclaimed. Staff also recommended adding materially that sentence to section 3. The checked numbered proposal rules do not show the reclaim sentence as an independently enacted numbered clause.

These are at least two consequence formulations in distinct textual locations: explanatory non-recognition plus return, and staff-assessed non-compliance plus reclaim.

Conflating them would create authority by grammar. It would turn a staff view and recommendation into an enacted confiscatory instruction merely because both appear on the same public page. Staff can identify operational concerns and recommend implementation language. Staff assessment is not automatically policy, a Board instrument, a court order or proof of lawful delegation. A serious account must therefore hold both facts at once: the reclaim prospect is real enough to require scrutiny, while its enacted status, location and lawful execution are not closed by the checked record.

In the normal route, D3 contemplates written approval before record completion; this article begins where parties are said to have acted outside that route.

That single sentence is enough context because the issue is not the ordinary completion chain. It is what should happen after a claim falls outside it. The answer cannot be “anything goes.” Accurate registration is important. A buyer should not be inserted into WHOIS or RDAP merely by sending an unverified assertion. Nor can the answer be “nothing outside our process has legal or practical existence.” The registry did not create every contract, every payment, every corporate authorisation, every insolvency instruction, every court-recognised interest or every operational fact that it may later be asked to record.

What AFRINIC announced, and what remains unproved

The dated public sequence is clear in some respects. D3 was submitted on 22 November 2021. The interval between submission and AFRINIC’s announced ratification on 4 February 2026 was 1,535 calendar days. That elapsed time is not itself an implementation period, and it is not proof of unlawful delay. An archived RPD message on 12 October 2025 described D3 as being at consensus and sent to the Board, with ratification still pending; the message is a participant’s archived statement, not an independent legal determination.

On 10 December 2025, the AFRINIC-36 policy liaison presentation still listed the proposal as at consensus and awaiting Board ratification.

AFRINIC then published an official overview dated 4 February 2026 naming AFPUB-2020-GEN-006-DRAFT03 as ratified. In that overview, AFRINIC said approved transfers are reflected in WHOIS and RDAP, and it contrasted controlled redistribution with informal or unregistered transfers. Those statements establish AFRINIC’s announcement and its account of the registry outcome. Its accompanying vocabulary of authoritative oversight, stewardship and benefit remains institutional self-description. It does not establish sovereign jurisdiction or a power to punish.

On 18 February, AFRINIC defended the process, describing the ratifying Board as duly constituted and invoking documented policy-development consensus, its Bylaws and stewardship. That is AFRINIC’s position. The checked public ratification material does not disclose a numbered Board resolution, meeting minutes, motion, vote, quorum, recusals, legal opinion or receiver direction for this act. Where the authority of the decision-makers is material, they are the individuals AFRINIC presents as its Board, not an uncontested legislature whose declarations bind every legal and operational reality.

AFRINIC’s court index records a plaint challenging ratification, Skyconnect v AFRINIC & Anor, filed on 9 March 2026 under the reference SC/COM/PWS/000132/2026. That was 33 calendar days after the announced ratification. The filing proves that a challenge exists; it proves neither the challenger’s case nor AFRINIC’s defence. No final merits disposition was located as of 10 August 2026. An earlier injunction application concerning a resource-transfer policy, dated 25 June 2021 in the later court index, was described there as not granted and withdrawn on 6 January 2022. That history did not adjudicate the merits of the 2026 ratification.

The implementation record is also incomplete. The D3 staff assessment anticipated hostmaster review, changes to forms and the portal, tagging, logs, agreement revision and coordination between registries. Yet no checked source proves that a reclaim was executed under D3 after the February 2026 announcement. The record does not close whether or where the relevant D3 language was incorporated into the current policy manual, whether the staff-recommended reclaim sentence became enforceable text, which official is delegated to issue a non-recognition, return or reclaim instruction, or what safeguards precede irreversible action.

These are not marginal drafting questions. A statement that “the policy was ratified” cannot answer which sentence was enacted, where it sits, who may apply it or how a person can challenge its use. A label at the top of a public page cannot carry all the legal and operational content missing beneath it.

The strongest case for refusing recognition

The strongest contrary case begins with the fragility of shared records. Suppose two private parties report inconsistent claims over the same address block. Suppose the purported transferor’s signatory lacks authority, the recipient identity is incomplete, consideration records do not match, the claim conflicts with a court hold, or a counterpart registry cannot confirm its side of an inter-registry update. If AFRINIC changed its public record immediately, it could present a false holder, enable a duplicate claim, detach contacts from the party still operating the network, or deepen inconsistencies in related systems.

A categorical warning against shadow transfers may also discourage parties from bypassing the evidence needed for a reliable common ledger. Users of WHOIS and RDAP need to know that the published holder field was not changed simply because one claimant reached the intake desk first. Incident responders need usable contacts. The registry needs an audit trail. Other relying parties need a stable last-known state while competing assertions are tested. A required return could be understood, in its least expansive form, as restoration of that last verifiable state while defects are cured.

This defence is powerful because coordination fails if a registry must record every asserted transaction. The doctrine does not require credulity. It permits AFRINIC to verify source authority, recipient identity, duplicate claims, fraud indicators, court holds and record integrity. Official guidance concerning legacy resource holders likewise describes an entitlement-review and record-update function. Evidence checking is a proper administrative task.

What the defence cannot support is the leap from “unverified” to “nonexistent,” or from “record conflict” to “confiscate.” The very risks that justify caution also justify precision. If the record cannot yet say that the claimant is the verified holder, it should not say that the claimant has no possible interest. If the parties’ documents are defective, the registry should identify the defect. If there are competing claims, it should preserve the conflict. If legal ownership or authority is disputed, a competent court or genuinely independent reviewer must decide that dispute within its jurisdiction.

The same institutional chain should not be claimant, rule-maker, fact-finder, final reviewer and executioner.

A ledger is a witness, not an author of reality

AFRINIC is a private regional number-registry service provider. Its legitimate control surface includes its request intake, evidence review, public registry state, service records and the changes it publishes. Within that surface, it may decide that evidence is insufficient for a clean holder update. It may preserve the last verified state, prevent duplicate registry entries, flag a conflict and maintain logs. It may protect continuity while verification proceeds.

Its authority stops at the boundary of its service. AFRINIC is not a sovereign, regulator, police force or punishment authority. It has no universal title jurisdiction. Number registration coordinates uniqueness; it does not confer on the coordinator the power to erase every transaction fact outside its interface. A database entry can be important evidence. It can be relied upon. It can shape access to associated services. None of those features turns the database into the author of all rights and conduct surrounding the resource.

This is why the difference between non-recognition and invalidity must stay visible. Non-recognition can accurately describe AFRINIC’s state of knowledge: it has not accepted a claimed holder change. Invalidity purports to settle a larger legal question: no transaction or interest exists. AFRINIC can make the first statement about its own records. It cannot manufacture the second as a universal conclusion simply by attaching the word “policy.”

An executed contract might be valid or invalid for reasons that a court can assess. Consideration might have changed hands. Corporate authority might be contested. An insolvency instrument might govern control. An operator might already announce the addresses. A prior holder might retain access to contacts or registry-linked functions. A court might later recognise one party’s interest. These realities can be inconsistent and uncomfortable, but discomfort does not authorise a bookkeeper to delete them conceptually. The correct record is sometimes a record of unresolved conflict.

The more economically important the registry interface becomes, the more disciplined this boundary must be. IPv4 scarcity has turned administrative discretion over records into practical power over valuable and operationally necessary resources. The person controlling the interface may influence financing, settlement, incident response, customer confidence and continuity without bearing the loss if the decision is wrong. That asymmetry does not make recordkeeping illegitimate. It makes procedural restraint essential.

From refusal to punishment

Non-recognition becomes enforcement creep when it is used as the first step of an adverse chain that the institution itself controls from allegation to irreversible outcome. The staff reclaim language brings that risk into view. Reclaim is not simply a neutral refusal to edit a row. In ordinary language and practical effect, reclaim points toward removal, reassignment or deprivation. When attached to a process violation rather than a proven record error, it can make the registry’s approval a condition of economic reality.

The shift can happen quietly. A holder is told that a transaction is not recognised. Associated services or records remain tied to the previous state. The parties face uncertainty over warranties, settlement and operational authority. If the next institutional step is required return or reclaim, the cost of disputing the bookkeeper’s conclusion rises sharply. Even without a proved outage or a completed reclaim case, the mechanism is visible: control of a critical interface creates a hold-up point.

The problem is not cured by calling the action stewardship. Official words such as community, consensus, authority, stability and benefit describe how AFRINIC presents its role. They do not supply treaty jurisdiction, statutory police power or a general mandate to confiscate. A private coordinator cannot launder a larger authority through repeated institutional vocabulary. Nor can dependence on its services convert acquiescence into meaningful consent to punishment.

The practical distinction can be stated as a sequence. First, AFRINIC receives an alleged outside-process transfer. Second, it checks whether the assertion is supported. Third, if material uncertainty remains, it preserves the last verified registry state and publishes a bounded conflict status. Fourth, it gives notice, states the missing evidence and allows cure. Fifth, a genuinely independent forum can review a materially adverse decision before an irreversible change. Only after reasoned resolution should the registry correct or preserve its state through a continuity-safe plan.

That is a minimally reviewable chain of at least five stages: notice, evidence and cure, provisional conflict state, independent review, and safe correction or preservation. It is a governance design, not a claim that AFRINIC currently provides all five. No checked source closes a dedicated outside-process notice form, conflict-state field, evidentiary standard, cure clock, interim continuity rule, independent appeal, injunction route, reclaim authorisation, reversal protocol or public completed example.

Without that chain, “return” is dangerously underspecified. Return to whom? On whose evidence? By which official? Is the last recorded holder necessarily the party with the strongest current legal claim? What happens if the prior holder has dissolved, entered insolvency or disputes the transaction? What associated registry state is preserved? Can the instruction be paused? Which reviewer can reverse it, and how is operational continuity protected while review occurs? The record supplied here does not answer those questions. It would be false to invent answers from the fact of ratification.

Conflict metadata is the truthful alternative

A conflict-state model begins by refusing a false binary. The registry need not choose immediately between accepting an unverified transfer as clean and treating it as though it never occurred. It can retain the last verified holder while adding a bounded indication that a competing claim has been submitted. The flag should identify the category of uncertainty without publishing unnecessary private material. It should record timestamps, affected registry functions, the existence of notices and the procedural status of review.

The purpose is not to create a permanent grey market inside the registry. It is to make uncertainty legible while the parties assemble proof. A conflict flag tells relying parties that the visible holder is the last verified state, not necessarily a final adjudication of every interest. That is more accurate than silently freezing an old record and more responsible than accepting a new claim without evidence.

The next element is an objective evidence list. The parties may need to show executed instruments, authority of signatories, relevant corporate records, consideration or settlement evidence, control evidence, court or insolvency instruments, and any material counterpart-registry confirmation. The registry’s inquiry must remain tied to its legitimate job: whether it can safely publish a holder state, maintain uniqueness, avoid duplicates, respect court holds and protect continuity. It may not expand into a general judgment about whether it approves the commercial bargain.

Notice turns that evidence list into due process rather than an invisible test. Every materially affected claimant should know that non-recognition is provisional, what evidence is missing, which facts are disputed, which records will remain unchanged and what actions are contemplated. A written reason matters because it constrains arbitrary variation between cases and gives a reviewer something concrete to assess.

Cure is equally important. If a missing authorisation, inconsistent identity record or incomplete counterpart confirmation can be fixed, the policy should make formal correction easier than concealment. A published clock is needed so “provisional” does not become indefinite. The available record does not establish the duration of any appropriate clock, so no number should be invented. The principle is that the deadline, extension grounds and consequences must be known before the adverse action, not improvised after a party objects.

Continuity must be separated from merits. While proof is contested, the registry should preserve the last verified operational state and avoid irreversible changes to WHOIS, RDAP, transfer logs, reverse DNS or RPKI-related authority without a safe plan. Preservation is not an award of ownership to the previously recorded party. It is an interim technique for reducing harm while the forum works. Essential records and notice trails must remain accessible to the affected parties and reviewer.

Independent review closes the chain. A reviewer outside the same management line should be able to inspect the evidence, require exchange, preserve the status quo where necessary, assess whether the registry stayed within its function and direct a clerical correction. Disputed legal rights that exceed the registry’s competence belong before a competent court. Review need not decide unrelated commercial issues. Its immediate purpose is to stop the registry from turning an unresolved record conflict into irreversible punishment by its own assertion.

Why operators cannot treat this as paperwork

For a buyer, the obvious exposure is a transaction that has economic and operational substance but remains absent from the registry. Payment or control may have moved while AFRINIC continues to record the seller. That creates a difficult settlement state: the buyer may be using the addresses while lacking the registry recognition on which customers, auditors or counterparties expect to rely. The mismatch can affect warranties, escrow release and the credibility of future transfers or disputes.

For a seller, the risk runs in both directions. It may believe it has transferred control yet remain visible as the holder. It may still receive incident contacts or face questions about conduct it no longer controls. It may also confront a buyer that alleges failure to deliver a registrable interest. A blanket non-recognition rule does not resolve those private obligations; it can intensify them by withholding a precise account of what evidence would cure the record.

Operational teams face a different layer of exposure. Stale contacts, reverse DNS, route objects or RPKI-related authority can complicate incident response during a record conflict. The checked evidence does not prove that D3 has caused a specific outage, routing incident or security failure. The risk mechanism is nevertheless concrete: the operational user and the registry-recognised holder can diverge, leaving authority and response paths inconsistent. A continuity rule should manage that divergence without presuming the merits.

Smaller operators bear the asymmetry more heavily. They are less able to fund specialist legal work, tolerate long uncertainty or obtain substitute address space. A process that is formally available but slow, opaque or dependent on the same decision chain can therefore function as a practical denial. No available dataset quantifies processing time, legal cost, price discount or denial rate, so the magnitude cannot be stated. The direction of exposure follows from the combination of scarce resources, concentrated registry control and weak liability for a mistaken adverse action.

Financing and customer confidence can also be affected. An unresolved registry mismatch may lead a lender, buyer or major customer to demand additional evidence or contractual protection. The correct analytical expression is not a fabricated price estimate. Expected registry-risk exposure consists of the probability of non-recognition or adverse action multiplied by the exposed asset or transaction value, plus documentation and legal cost, plus operational continuity cost. The evidence supplies none of the numerical inputs needed to calculate a result.

The deepest economic risk is behavioural. If presenting a disputed transaction to the registry exposes the parties to automatic return or reclaim, they have an incentive to hide rather than correct it. They may rely on leases, nominees, incomplete disclosures or parallel evidence of control. That makes the common registry less aligned with commercial and operational reality. A policy intended to protect accuracy can therefore degrade accuracy if its remedy is punitive.

NRS treats the policy direction as a lock-in and continuity risk requiring direct operator and member action. Its proper role is to advocate, research, convene and represent expressly authorised members. It does not process transfers, run WHOIS or RDAP, operate RPKI, adjudicate disputes, hold assets or guarantee continuity. LARUS analysis identifies selective registry recognition, mandate laundering and dependence on the registry interface as structural asset risks. BTW research supplies the mechanisms of capital control, registry-layer risk and independent review.

These first-class materials establish the institutional design concern without turning any advocate into the substitute registry or court.

The record should show its own limits

A trustworthy registry does not have to pretend omniscience. It earns trust by distinguishing a verified state from a provisional state, an allegation from a finding, and a clerical decision from an adjudication. The proposal’s language becomes most dangerous where those categories are compressed into the single word “recognise.”

Recognition can mean that AFRINIC has accepted evidence sufficient for its database. It must not be allowed to imply that AFRINIC has conferred existence on every right. Return can mean correcting a record to the last supported state. It must not be allowed to imply confiscation by a private institution. Reclaim can describe a staff recommendation. It must not be reported as a separately enacted sanction without the missing textual and authority chain.

The immediate demand is therefore modest but exacting. Publish the precise ratification instrument. Identify the enacted location and status of the non-recognition, return and reclaim formulations. State who is delegated to act. Provide the evidence checklist, notice method, cure path, interim continuity rule, reasoned decision and independent review route. Disclose a completed-case or audit record when one exists. Do not treat the absence of those materials as permission to fill the gaps with institutional claims.

Until that record exists, AFRINIC can still perform the legitimate part of its job. It can preserve the last verified entry. It can refuse to publish an unsupported holder change. It can document a bounded conflict. It can protect uniqueness and associated continuity. What it cannot do is use uncertainty as a doorway to punishment.