Summary

  • AFRINIC’s D3 proposal summary says number resources are non-transferable unless AFRINIC expressly approves a transfer request in writing, while AFRINIC’s ratification overview presents the update of WHOIS/RDAP as the completion path for approved transfers. That language describes completion within AFRINIC’s recognition and record system, not a universal judgment about every contract, beneficial interest, corporate act or operational fact.
  • Three realities must remain separate: private parties can make a transaction under applicable law; AFRINIC can recognise or withhold recognition after checking the instruction; and AFRINIC can change its database. A private agreement does not automatically rewrite WHOIS/RDAP, yet a database update does not create all legal and economic reality.
  • Accurate verification and recording are indispensable bookkeeper functions. AFRINIC may authenticate control, prevent double recording, identify fraud or conflict, coordinate a safe transition and maintain continuity. It has no sovereign, punitive or commercial-licensing authority to make lawful transaction reality depend on discretionary permission.
  • The checked public record does not disclose the written-approval form, the exact signatory or delegation, an objective service deadline, a complete decision log, the incorporated policy text that closes the approval sentence’s legal force, or a completed transfer under the ratified D3 process. AFRINIC’s claim that its purported Board was duly constituted is an institutional assertion, while a recorded court challenge remains unresolved on the merits.

The transaction with three closing times

A transfer can appear to have one closing date and still contain three different events. The parties may execute documents, pay consideration, reorganise a company, transfer contractual or beneficial interests, or change operational control under the law that governs them. AFRINIC may then decide whether it will recognise the instruction as fit for its registry. Its staff may later alter WHOIS/RDAP and related transfer records. Those events can be coordinated, but they are not identical merely because one institution prefers to call the last event “completion.”

The distinction is not semantic housekeeping. It determines who possesses what kind of power. The parties and any competent legal authority govern the underlying transaction and its rights. AFRINIC governs the accuracy of its own registry service. A registry update changes the public record AFRINIC maintains and the party it recognises for service purposes. It can affect operational confidence, contact data and the safe coordination of related records. It does not, by the force of a database entry alone, originate every contract, erase every pre-existing interest or pronounce a universal legal title.

The reverse distinction matters just as much. A buyer and seller cannot type a contract and expect AFRINIC’s database to rewrite itself. The registry must receive an instruction, authenticate the source, identify the resource, check the recipient data, detect duplicate or conflicting claims and carry out a controlled change. For an inter-registry movement, a counterpart registry must coordinate its side. The old and new records must not both present incompatible certainty. Reverse-DNS and security continuity may also need careful handling, although the exact D3 sequence for those systems was not disclosed in the checked record.

There is therefore a legitimate completion that belongs to AFRINIC: completion of AFRINIC’s verification, recognition and recording work. That completion is valuable. It creates a dependable published state and an audit point on which operators and counterparties can rely. But it is narrower than the completion of all transaction reality. The title of the policy architecture can be true inside the registry process and false as a universal proposition. “No transfer is complete” must mean “no transfer is complete in AFRINIC’s recognised records” rather than “no legally or economically meaningful change exists anywhere until AFRINIC permits it.”

This boundary is the whole contest. If written approval authorises staff to make a safe record change, it is a procedural control. If it purports to authorise the parties’ commercial act itself, it becomes a licence. The same sheet of paper can perform either role depending on its grounds, consequences, timing and review. The database interface gives AFRINIC the practical ability to delay recognition. It does not supply sovereign jurisdiction over the transaction that awaits recognition.

What the ratified material actually says

The D3 proposal was submitted on 22 November 2021. Its explanatory summary says number resources are non-transferable unless AFRINIC has expressly and in writing approved a transfer request, and it assigns AFRINIC the task of making prudent approval decisions. The proposal also says AFRINIC will process and record qualifying inter-registry transfers. That is the textual source of the two-step institutional chain: first the written decision, then the record action.

Precision about the location of the language is essential. The express-written-approval sentence appears in the proposal’s summary of how the proposal addresses the stated problem. It is not presented as a separately numbered rule in proposal sections 3.1 through 3.6. The numbered material sets out source, recipient, category and recording conditions, but the checked record does not provide an incorporated policy-manual passage that conclusively resolves the normative force of the summary sentence. The condition should neither be dismissed nor promoted into a clause that the published text does not show.

On 10 December 2025, an AFRINIC-36 policy presentation described D3 as having consensus and awaiting ratification. AFRINIC’s official overview dated 4 February 2026 names D3 as ratified and says all approved transfers are reflected in its WHOIS/RDAP registry database. AFRINIC’s communiqué of 18 February says its purported Board ratified the proposal after what the institution describes as documented policy-development consensus. These records establish what AFRINIC published, the sequence it announced and the authority it claimed.

They do not independently prove that consensus delegated sovereign power, that the purported Board’s constitution was legally uncontested or that every contemplated operational step was live.

AFRINIC’s legacy-resource guidance supplies a useful description of its operational thinking. AFRINIC says it verifies evidence of entitlement before permitting registration changes, requires policy conditions to be met before a resource is transferred, and updates organisation and number-resource records after it determines entitlement and receives the required information. The guidance connects institutional due diligence, completion, status consequences and a database update. It does not establish that AFRINIC is a party to every private transaction or the creator of every underlying interest.

The official materials therefore prove a rule text, a ratification announcement and AFRINIC’s description of its intended record consequences. They do not prove the legitimacy of every authority claim embedded in AFRINIC’s language. References to mandate, stewardship, oversight, consensus or an authoritative registry remain AFRINIC’s self-description. The fact that the institution publishes the principal record explains why accuracy matters; it does not transform operational centrality into political sovereignty.

The implementation evidence is also incomplete. The D3 staff assessment anticipated resource-intensive hostmaster evaluation, portal work, forms, resource tagging, transfer-log changes, agreement revision and coordination with another registry. That is evidence that approval is not a purely ceremonial signature. It requires people, systems, instructions and transitions. Yet no checked public source by 10 August 2026 closed the actual implementation date, a current approval form, the identity or delegation of the deciding signatory, a defined service deadline, or a completed transaction under the ratified D3 process.

The architecture is visible; its fully operating machinery is not.

The governing bookkeeper doctrine

Heng Lu doctrine supplies the non-negotiable control truth: a registry record describes reality; it does not create it. A number registry is a bookkeeper and coordinator. Its legitimate work is to maintain accurate records, verify control, protect uniqueness, prevent fraud and double recording, isolate conflicts, and preserve reverse-DNS and security continuity. Those tasks are serious precisely because other actors rely on the ledger. They do not include sovereignty, punishment or commercial permission.

The doctrine does not trivialise the database. Calling AFRINIC a bookkeeper is not an insult and does not mean anyone may ignore registration. A reliable bookkeeper holds an essential interface. A stale or contradictory public record can misdirect communications, complicate incident response and leave operators uncertain about who is authorised to request changes. A careless update can facilitate fraud or produce two competing claims to one resource. Strong bookkeeping is a public operational good.

But the value of the interface cannot be used to reverse the relationship between fact and record. When a company acquires contractual or beneficial interests under applicable law, the database should be brought into alignment after objective verification. The update is evidence of recognition and an important service event. It is not the metaphysical origin of the acquisition. If a competent court orders a hold or resolves a dispute, that legal act is categorically different from an internal registry decision.

AFRINIC can reflect the legal result and protect the record while it is contested; it cannot replace the independent source of binding law.

This is where written approval becomes dangerous. In its lawful form, the approval says: the instruction has been authenticated, the resource is correctly identified, the source has recognised authority, the recipient data are sufficient, no duplicate or disabling conflict is present, and staff may safely alter the ledger. In its inflated form, the approval says: AFRINIC has decided that the underlying bargain is commercially, geographically or politically acceptable, and without that permission the parties’ reality is void. The first statement is bookkeeping. The second is unmandated rule over assets and agreements.

AFRINIC has no treaty, statutory delegation or sovereign jurisdiction established in the checked material that would justify the second statement. Its control of a private database is not a source of ownership. Community process does not manufacture public-law authority. A vote by a purported Board cannot enlarge a coordination function into a general licence over private economic acts merely by using the word “approval.” Institutional procedure may govern the institution; it does not automatically bind the rest of reality as sovereign law.

Written approval must therefore be interpreted and designed narrowly. It can authorise AFRINIC’s own action. It can confirm that specified evidence satisfies published, objective record-integrity conditions. It can explain why an update is safe or why a defect prevents it. It cannot be a blank instrument for judging price, business model, commercial need, political acceptability or other matters that do not bear on the integrity of the record. The burden belongs on AFRINIC to show that every requested fact serves proof, uniqueness, fraud prevention, conflict management or operational continuity.

Why the database still matters

The strongest case for AFRINIC’s completion language begins with a practical truth: a unique number-resource ledger cannot update itself. If a transfer were treated as operationally settled in every respect before the registry authenticated the source and changed the public record, stale contacts, incompatible records, confused authority and fraud would become more likely. A written decision can give the parties a clear condition for coordinating payment, record change and the operational handoff. It can protect both sides from a closing in which one side performs and the essential record never moves.

That case is powerful because the registry is not just an archive of old facts. It is a live coordination service. Operators, incident responders, counterpart registries and commercial actors may consult its state. The transfer log and WHOIS/RDAP entries can make a change legible. Where a related transition touches reverse DNS or RPKI, continuity must be managed rather than assumed, even though the exact order contemplated for D3 remains unknown. A registry that changed on an unauthenticated email would be useless and dangerous.

The doctrine accepts all of this. It simply refuses the unwarranted final step. The need to authenticate a record instruction does not make AFRINIC the creator of the contract. The need to prevent duplicate entries does not authorise it to judge the parties’ price. The value of a coordinated closing does not make the registry a beneficiary of the bargain. The possibility of fraud justifies proof; it does not justify open-ended discretion. The database matters enough to require disciplined process, not enough to become a throne.

The proper response to a gap between asserted transaction reality and verified registry state is truthful metadata. If the instruction is incomplete, AFRINIC can preserve the last verified entry while displaying a bounded pending or conflict state, where safe and appropriate. It can identify the missing proof, maintain continuity and route a rights dispute to independent review or a competent court. What it should not do is force the ledger to proclaim that nothing exists outside the last unchallenged row.

A record can accurately say that recognition is pending without pretending that no agreement, payment, corporate act or operational change has occurred.

This approach protects the informational value of WHOIS/RDAP. When recording is predictable, timely and non-punitive, parties have a reason to reveal the facts needed to align the database. When approaching the registry exposes them to an opaque veto, delay or collateral demands, they have an incentive to arrange economic control without asking for a record change. Nominee, lease or other indirect structures can then separate the public ledger from underlying control. The result is the opposite of the transparency the approval requirement is supposed to produce.

Accuracy therefore depends on institutional restraint. The registry earns reliable information by making safe corrections easy. It degrades its own data when recognition becomes a weapon. A narrow approval process reinforces the ledger because it is trusted as a verification service. An expansive approval monopoly weakens the ledger because parties treat it as a risk to be routed around.

The approval-and-update gap

The policy architecture contains at least two AFRINIC-controlled transitions. First, someone must issue or communicate the written approval. Second, registry operations must implement the update. In a coordinated inter-registry case, another registry has work on its side as well. This count does not establish a fixed number of forms, staff touches or days. It identifies the minimum institutional handoffs on which completion in AFRINIC’s system depends.

Those handoffs create timing risk even if every request is eventually accepted. A buyer may commit capital before the record changes. A seller may be reluctant to release operational control before the approval is documented. Escrow instructions may need to define whether release occurs on approval, on a visible WHOIS/RDAP update, on completion of related record transitions, or on some combination. If the signatory, decision clock and implementation sequence are unpublished, the parties cannot price the duration of their exposure with confidence.

The economic mechanism is straightforward even though its magnitude is unknown. Expected hold-up cost consists of the probability of delay or rejection multiplied by the exposed transaction value, plus documentation, legal and operational transition costs. The record contains no measured probability, transfer value, processing-cost dataset, denial rate or average delay for D3, so no numerical result is defensible. But the direction of risk does not require an invented statistic.

When one actor controls the last necessary interface and its decision conditions are opaque, counterparties must reserve time, capital and contingency for that uncertainty.

Small operators are especially sensitive to the structure, not because the evidence supplies a measured distribution, but because the ability to carry delayed capital and retain specialist advisers is uneven. A large acquirer may absorb another review cycle. A smaller network may have a narrower financing window or fewer people able to answer repeated documentary requests. The same nominal rule can therefore create unequal practical bargaining power without ever announcing a discriminatory outcome.

Continuity is another part of the gap. The party exercising operational control may change before AFRINIC’s public contacts do. Incident responders may then see stale information. The source may remain the recognised contact while the recipient expects to direct the resource. Reverse-DNS or security authority could require a coordinated transition, but the public evidence does not establish the exact D3 sequence. The answer is a documented handoff and truthful interim status, not the fiction that whichever row remains visible contains the entirety of reality.

Financing also turns on recognition risk. A lender or acquirer evaluating a resource-related transaction will care whether registry recognition is objective, time-bound and reviewable. Unmeasured timing is not the same as zero cost. An opaque approval condition can invite a discount or additional closing protections because the capital provider cannot know when the record will settle. Again, the checked evidence supplies no percentage discount and none should be invented. The analytical point is that uncertainty at a necessary interface enters the transaction even when the registry claims no ownership.

Authority without a disclosed chain

The 4 February 2026 announcement does not close the authority chain behind the approval gate. The public ratification material checked for this analysis does not disclose a numbered resolution, meeting minutes, the motion, vote, quorum, recusals, a supporting legal opinion or a receiver direction for this act. AFRINIC says the ratifying body was duly constituted. Where authority matters, it must be described as AFRINIC’s purported Board because the institutional label is contested rather than independently established by the announcement itself.

AFRINIC’s court-case index records a plaint filed on 9 March 2026 challenging the ratification, 33 calendar days after the announced date. The existence of the filing establishes a live challenge in the public record. It does not establish that the ratification was valid or invalid. No final merits judgment was located by the evidence cutoff, so neither side may convert a docket entry into a judicial conclusion.

The distinction between institutional action and authority is crucial. AFRINIC can prove that it published a ratification overview and defended the process. It can prove what its proposal says and what its systems may be asked to do. Those acts do not, without more, prove that the purported Board possessed the power to impose an irreversible commercial licence. Even a perfectly documented internal vote would govern the organisation only within the legitimate scope of the organisation’s function. Procedure cannot launder a missing sovereign mandate.

The receiver boundary remains open as well. No checked source discloses a receiver instruction that closes ratification or implementation authority. A receiver’s stabilisation or election function, if any, would not silently confer power to legislate permanent transaction permission. Any such direction would need to be separately evidenced and confined to its legal source. Silence cannot serve as delegation.

There is a second missing chain below the purported Board: the person or office that gives the written approval. The public record does not identify the exact signatory, publish the delegation instrument, provide a decision template or establish an objective service level. A written approval with no disclosed author, clock, criteria or review route is not a transparent control. It is a chokepoint whose legal and operational boundaries cannot be audited from the published material.

The legitimate version would be simple to recognise. A dated decision would identify the request and resource, name the decision-maker and delegated authority, state the objective evidence considered, list any defect, provide a cure period, explain the consequence for the record and identify an independent review route. The implementation log would then show when WHOIS/RDAP and related records changed. Each step would be visible as an act of registry administration rather than an unexplained licence over the bargain.

A proportionate completion rule

AFRINIC may properly refuse to alter its database when the instruction is fraudulent, duplicates a claim, comes from an unauthorised representative or conflicts with a competent court hold. It may require reliable source and recipient identification. It may coordinate with a counterpart registry so that two ledgers do not settle inconsistently. It may preserve the last verified state while a material conflict is reviewed. Each ground follows from the duty to maintain a unique, accurate and safe record.

The approval decision should stop there. A registry-integrity test asks whether the proposed entry is accurate and safe. A commercial-licensing test asks whether the institution approves of the transaction itself. The former can be answered with documents and enumerated defects. The latter invites shifting ideas of need, fairness, policy preference or institutional leverage. AFRINIC’s database control supplies no lawful basis for the latter inquiry.

Notice, reasons, cure and review convert discretion into accountable administration. Notice tells the parties what the registry believes is missing. Reasons connect a refusal to a published criterion. Cure permits a correctable documentary problem to be solved without destroying the transaction. Independent review separates the person who made the initial decision from the authority that assesses a genuine dispute. Interim continuity measures prevent a process disagreement from becoming operational damage.

A published service period is equally important. The checked record does not reveal one, so this is a design requirement rather than a description of current performance. The clock should define when a complete request is acknowledged, when defects are notified, which events suspend the period and when the record will change after approval. Without those points, “written approval” is not a closing condition that parties can reliably coordinate; it is an indefinite dependency.

Simultaneous or escrowed settlement can align commercial performance with registry work without giving AFRINIC a stake in the price or transaction. Funds or other consideration can remain conditional while objective record checks proceed. Release can be tied to the agreed evidence of registry settlement. AFRINIC’s role remains authentication and recording. It need not become buyer, seller, beneficiary, valuation authority or commercial judge.

Where proof is incomplete or contested, a conflict-state record offers a better answer than false certainty. AFRINIC can preserve the last verified entry, display an appropriately bounded flag, protect continuity and identify the forum that can decide the disputed rights. This does not mean that any assertion should overwrite the registry. It means the registry can represent uncertainty honestly. A database that distinguishes verified state from pending claims is more accurate than one that treats institutional silence as proof that no external reality exists.

The completion rule should therefore read in substance as a limit on AFRINIC, not a grant of power over everyone else: AFRINIC’s record change is complete only after it has objectively verified the instruction and updated the relevant registry state. The rule should not declare that all aspects of the parties’ transaction are nonexistent before that moment. This formulation gives the database the attention it needs while keeping the bookkeeper within its mandate.

What remains unknown

Several gaps prevent confident claims about live operation. The published material does not close whether and where D3 has been incorporated into the current policy manual. It does not establish an implementation date or identify a completed D3 transfer. It does not disclose the approval form, signatory, delegation, fee schedule, service deadline, suspension clock or escalation route. It does not show the exact order of WHOIS/RDAP, transfer-log, reverse-DNS and RPKI changes.

The governance record is also incomplete. The purported Board’s numbered ratification instrument, minutes, motion, quorum, vote and recusals were not disclosed in the checked sources. Any receiver role remains unproven. The March 2026 litigation is a challenge, not a merits decision. These uncertainties do not erase the policy statement AFRINIC published; they limit what can be claimed about its authority and implementation.

No public performance dataset in the checked record measures volume, processing time, denial, withdrawal, transaction cost or price effects under D3. Delay and hold-up can therefore be analysed as mechanisms, not reported as observed averages. An operator should not infer that the absence of a measured delay proves instant processing, just as a critic should not invent a delay to dramatise the risk.

The unresolved evidence also defines a practical accountability agenda. AFRINIC can close much of the uncertainty by publishing the exact ratification instrument, implementation delegation, approval criteria, complete request form, service clocks, reasons format, audit trail and independent review path. It can publish anonymised performance measures once transactions exist. Transparency would not answer every legal question, but it would reveal whether written approval operates as disciplined verification or discretionary permission.

Until then, the correct conclusion is bounded. AFRINIC announced a policy architecture in which its written approval and record update complete a transfer for its own recognition system. Private agreement alone does not update that system. The system’s update alone does not create all the facts and rights outside it. The bookkeeper must be exact, fast and reviewable because the ledger matters; it must remain thin because the ledger is not sovereign.