Summary

  • AFRINIC declared on 4 February 2026 that its Board had ratified AFPUB-2020-GEN-006-DRAFT03, a transfer proposal whose co-chairs had upheld consensus in January 2022. The later policy appeal was rejected as inadmissible, not decided on its merits.
  • The declaration does not collapse four different questions into one answer: the PDP record, the corporate act, the act’s judicial legitimacy and operational implementation each have a different evidential status.
  • D3 would classify resources by origin. AFRINIC-pool resources become Regional and cannot be transferred out of the region by an AFRINIC member; inbound resources become Global and may move under compatible inter-registry arrangements. That distinction turns registry provenance into economic exit control.
  • The strongest defence is real: a trustworthy transfer ledger can prevent duplicate claims, fraud and corrupted registration data. But those necessities do not establish that AFRINIC needs power over geography, business need, market exit or reclamation. At the evidence cutoff, the authority chain remained contested and completed implementation remained unproven.

Analysis

One declaration, four unfinished states

The most important fact about AFRINIC’s 4 February act is also the easiest to lose in institutional shorthand. “Ratified” describes what AFRINIC declared its Board had done. It is strong evidence of an official corporate claim and of the policy text the institution intended to advance. It is not a universal solvent. It does not by itself prove that every PDP objection was resolved on the merits, that the acting directors’ mandate was beyond judicial dispute, that a court-appointed receiver had been discharged, that systems had been changed, or that a transfer had been completed under the new rules.

Those are four states, not four phrasings of the same state.

First, there is the PDP state. Version 3 of the Number Resources Transfer Policy entered Last Call in December 2021. Co-chairs Vincent Ngundi and Darwin Da Costa announced on 14 January 2022 that objections were not new or had been addressed, upheld consensus and said the report would go to the Board. Mathanya Ramaboea, supported by three others, appealed that consensus determination on the same day. The Appeal Committee later set the appeal aside because it found the admissibility condition unmet: the appellant had not shown an effort to discuss and resolve the matter with the co-chairs or the working group. The committee did not adjudicate the substantive objections. “The appeal failed” is therefore accurate only if immediately qualified. It failed at the threshold. It did not produce a merits endorsement of D3.

Second, there is the corporate state. AFRINIC’s 2020 Bylaws say policies discussed and agreed through the Policy Development Process are ratified by the Board. They also define the Board as directors, not fewer than the required quorum, acting together. AFRINIC’s ratification overview says its Board ratified D3. Yet no separately published resolution, agenda paper, attendance record, quorum certificate, director-by-director vote, conflict record, legal opinion on the receivership transition or implementation authorisation was located by the 10 August 2026 cutoff. The absence of those public records does not prove that no valid meeting occurred. It does mean the public declaration cannot independently demonstrate the formal elements its own corporate definition makes relevant.

Third, there is the judicial state. AFRINIC announced eight elected directors in September 2025 after a rerun election. That proves the announcement and the institution’s position; it does not decide litigation over the election or the receiver’s authority. Receiver Gowtamsingh Dabee applied for termination or discharge of the receivership on 8 October 2025. Later AFRINIC communications said a decision was awaited and that the receiver would continue supporting the institution pending formal discharge. The maintained case list still marked the discharge matter and election-related proceedings ongoing at the cutoff. It also recorded a March 2026 plaint by Skyconnect, formerly Skyvision Guinee SA, specifically challenging the Board’s ratification of D3. That filing proves a live challenge, not that the challenge is correct. A related interim application was declined and not served on AFRINIC; an appeal concerning that order remained recorded. Neither the declined interim application nor the active main plaint supplies a final merits judgment.

Fourth, there is the operational state. D3’s staff assessment estimated implementation would take 12 months and require resource tagging, tools, system changes and additional Hostmaster work. The proposal page later carried a partial compatibility assessment: APNIC was described as reciprocal and bidirectional; ARIN as not reciprocal because of D3’s outbound restrictions; LACNIC compatibility as needing confirmation; and RIPE NCC as not requiring reciprocity. That matrix indicates implementation analysis, not a live production channel. At the cutoff, AFRINIC’s public Consolidated Policy Manual still described the earlier section 5.7 regime for IPv4 transfers within the AFRINIC region. Its current transfer guide likewise described intra-region IPv4 transfers. No implementation circular, completed D3 transfer, updated manual insertion, comprehensive compatibility map or first-transfer record was located.

The responsible conclusion is narrow. Ratification was declared. The PDP had reached a co-chair consensus determination. Corporate proof visible to the public was incomplete. Judicial legitimacy remained contested rather than decided. Operational completion was not publicly demonstrated. Any account that compresses those states into “the transfer policy is now fully in force” outruns the record. Any account that compresses them into “the policy is already void” does the same from the opposite direction.

How a 2021 draft reached a 2026 Board

The proposal did not begin during the 2025 governance transition. Version 1 was submitted on 17 October 2020. It sought to create one architecture spanning transfers within the AFRINIC service region, legacy resources and transfers involving another regional registry. It also introduced resource categories based on provenance. Version 2 followed on 8 October 2021 with revised definitions, references and language. Version 3, submitted on 22 November 2021 by Gregoire Olaotan Ehoumi, Noah Maina and Adeola A. P. Aina, is the text AFRINIC later marked ratified.

After discussion and the AFRINIC-34 policy meeting, the co-chairs sent D3 to Last Call from 8 December 2021 through 5 January 2022. The mailing-list record preserved objections. Among them were concerns that the draft left AFRINIC broad approval discretion and that a consequential transfer architecture required further scrutiny. On 14 January the co-chairs held that the objections were either not new or had been addressed and upheld consensus.

The appeal lodged that day matters because of what happened next—and what did not. In July 2022, the Appeal Committee concluded that the appellant had not satisfied a procedural prerequisite. It set the appeal aside as inadmissible. It did not test the disputed policy design, weigh the claimed defects or decide that open-ended discretion was justified. A threshold result may end an appeal without resolving the disagreement beneath it. That is precisely why later institutional references to a completed PDP must be distinguished from a merits adjudication.

D3 then remained awaiting Board ratification through AFRINIC’s governance interruption. A policy-liaison presentation in December 2025 still described it as having consensus and awaiting ratification. The presentation also referred to 2021 litigation over a different proposal, AFPUB-2019-V4-003-D4. That older case must not be reassigned to D3. The judicial challenge specifically concerning D3’s ratification came only after the 4 February 2026 declaration.

AFRINIC’s account is that functional governance structures were restored, the working group was reactivated and proposals with documented consensus could finally be considered. That is a coherent institutional narrative: an interrupted queue resumed after governance capacity returned. But “capacity returned” is the institution’s characterisation of a transition still entangled with an undischarged receiver and live proceedings. The passage of time did not refresh consent, answer objections or prove corporate authority. It merely carried the 2022 procedural state forward until the announced 2025 Board acted on it.

The timing makes the ratification a governance event, not merely a policy-administration event. The first consequential question is not whether transfer policy was needed in the abstract. It is why the acting Board chose to convert this particular dormant proposal into a ratified rule while its own authority chain and the receiver’s exit remained publicly unresolved. The next question is what that proposal actually empowers AFRINIC to do.

The origin tag that governs the exit

D3 defines scarce number resources as IPv4 address space and autonomous system numbers. It then creates four categories: Regional, Reserved, Legacy and Global. The labels appear technical, but their effect is commercial and institutional.

Resources in the AFRINIC pool—including IPv4 and ASNs received from IANA, recovered resources and certain resources received through the early-registration transfer process—are designated Regional. Special-purpose resources are Reserved. Historical legacy resources are Legacy. Resources transferred into AFRINIC from another registry become Global.

Within AFRINIC, Regional, Global and Legacy resources may be transferred. A transferred Legacy resource becomes Global. For an outward transfer, however, the origin category is decisive: an AFRINIC member may transfer out of the service region only a Global resource. A legacy holder may transfer a Legacy resource outward. The policy permits inbound resources from elsewhere and marks them Global. Reserved resources cannot ordinarily be transferred except through mergers, acquisitions and takeovers.

The architecture therefore does not merely ask whether a source controls a block, whether a recipient is identifiable or whether registry records can be accurately updated. It asks where the block entered the system. Two technically equivalent blocks can face different exit options because one originated in the AFRINIC pool and the other entered through an inter-registry transfer. The classification does not, on the available record, alter routing or the holder’s daily technical use. It governs transfer eligibility.

But transfer eligibility is a component of asset mobility, financing and corporate optionality even where packets continue to flow exactly as before.

D3 sets a /24 minimum for IPv4 transfers. It requires the source to demonstrate title or control and to declare no known or contemplated dispute. An AFRINIC recipient must enter the Registration Service Agreement, submit a detailed use plan and, if already holding resources, provide utilisation and compliance evidence. Inter-registry transfers depend on compatibility or reciprocity with the counterpart registry. Unauthorised transfers would be considered non-compliant and, according to the staff assessment, subject to reclamation.

Some of those controls serve legitimate ledger functions. Proof of control helps prevent double sale. A dispute flag can stop a contested block from being reassigned while ownership is examined. Accurate contacts and registry-to-registry acknowledgement keep WHOIS or RDAP records useful. Clear transaction logging helps operators, security teams and counterpart registries understand which party is responsible. None of that is trivial.

But other controls reach beyond keeping one truthful record. Origin-based outbound prohibition limits the set of acceptable registries and potential buyers. Recipient needs review lets the registry examine business deployment after a scarce resource has already been issued and made transferable. Non-recognition backed by reclamation can turn a record-accuracy function into a punishment surface. Compatibility decisions can become discretionary bottlenecks if criteria, service levels and appeal routes are not published.

D3’s own policy mirror reveals an institution moving between two roles: the indispensable bookkeeper that prevents contradictory records, and the gatekeeper that claims authority over where an asset may go.

The authority question the word “community” cannot answer

AFRINIC says D3 followed a bottom-up process. Its February response to member criticism invoked Bylaws articles 11.2 and 11.3, called the Board properly constituted and characterised attacks on the policy as misinformation. Those statements are evidence of AFRINIC’s legal and institutional position. They are not independent adjudication of that position.

Under the controlling Heng Lu doctrine, an RIR service region is an administrative scope, not sovereign territory. An open policy forum is a useful coordination mechanism, not a legislature. Participation can help discover workable rules. Consensus can tell an organisation whether active participants can live with a proposal. Neither turns participants into owners of every affected holder’s operational assets, and neither grants a private corporation territorial authority over value movement by people who did not authorise it.

That doctrine does not deny the need for a registry. It narrows the registry’s legitimate mandate to what running code and truthful coordination require: uniqueness, proof of control, accurate contact and security metadata, isolation of exact disputes, recording of completed transfers and continuity when institutional components fail. It rejects the inference that because a registry must record a transfer, it may also decide whether the holder’s commercial destination serves regional policy.

The distinction is sharper in a governance crisis. Concentrating policy creation, member contracts, registry records, approval and reclamation in a fragile private institution creates a single control point. The institution can make its own continuity a reason to restrict member exit. D3’s financial assessment said keeping AFRINIC-pool resources in the region prevents the loss of resource members to other RIRs and would have minimal revenue impact. That statement exposes the agency conflict. The party designing the exit rule has an institutional interest in retaining the accounts whose exit it regulates.

Resource holders bear any liquidity or collateral discount; registry decision-makers do not necessarily carry that loss on their own balance sheets.

Community rhetoric does not cure that conflict. The “AFRINIC community” is not the owner of member-held operational assets. Nor can claims of stewardship answer corporate-law questions about directors, quorum or receivership. It is possible for a policy process to be procedurally open and still claim too much substantive power. It is also possible for a good policy idea to be ratified by a body whose authority is disputed. Procedure, substance and authority must each stand on their own evidence.

The strongest case for D3, stated before the answer

The strongest defence of D3 begins with scarcity and disorder, not institutional ambition. IPv4 space is finite, and historical holdings are fragmented. Operators still need addresses while free pools have tightened. Without a documented transfer channel, commercial changes can occur without corresponding registry changes. Stale registration data can misidentify the responsible party. Competing claims can facilitate double sale. Weak validation can increase fraud, hijack risk and operational confusion.

D3 opens channels that AFRINIC’s older public policy did not provide. It can enable legacy-to-member transfers, certain inter-registry transfers and a common treatment of IPv4 resources and ASNs. A source-control test, dispute-free declaration, recipient agreement and coordinated updates can reduce ambiguity. A /24 minimum can avoid impractically small units. Reciprocal exchange can give counterpart registries assurance that records will align at both ends.

The regional-retention argument also deserves its serious form. African networks still face unmet demand. Wealthier foreign buyers may be able to outbid local operators. If all AFRINIC-origin space can immediately move to any market, supporters may fear a one-way export from a region that still needs digital infrastructure. A needs plan could be presented as a brake on speculation. A regional classification could be presented as stewardship of scarce capacity assigned for African development rather than as a market embargo.

The governance defence is equally intelligible. The Bylaws expressly include Board ratification. Co-chairs upheld consensus. The appeal did not proceed. After a period in which AFRINIC could not complete ordinary governance work, a new Board was announced and a dormant queue remained. Clearing that queue can be described as institutional recovery, not a power grab. The Board, receiver and staff may also have legitimate reasons not to publish privileged legal advice or transaction-sensitive implementation details while litigation is ongoing.

That is the best case because it recognises actual operational risks and an explicit corporate text. The answer cannot simply be that every transfer check is illegitimate or that any confidentiality proves misconduct. A registry that records anything at all must validate something. A court-appointed receiver may have to preserve continuity while corporate organs are restored. Pending litigation should not be converted into presumed guilt.

Why the defence does not establish the claimed power

The defence proves the need for a reliable ledger. It does not prove the need for an origin-based capital boundary.

Fraud prevention requires proof that the source controls the resource, not proof that a buyer remains within a region. Duplicate-claim prevention requires a single authoritative update or coordinated registry acknowledgements, not a prohibition on a particular destination. Dispute isolation requires an exact freeze attached to an identified block and parties, not general power to examine every recipient’s business plan. Accurate WHOIS or RDAP requires truthful contact and status data, not a registry veto over asset geography.

Routing-security continuity requires careful transition of operational metadata, not automatic reclamation when a private transaction falls outside the approved path.

The development argument also contains an unproved leap. Restricting outward sales could retain address inventory in the regional registration domain, but it could also reduce the collateral value and exit options of African-held assets. A holder expecting future lock-in may demand a discount before bringing Global resources into AFRINIC. Capital may prefer leasing, corporate wrappers or informal arrangements over a transparent sale. A smaller permitted buyer set can reduce price discovery and extend transaction time.

None of those effects has a precise AFRINIC-specific measurement in the source record, but each follows from the structure of constrained exit. “Keep addresses in Africa” cannot simply be assumed to mean “improve African connectivity.” The policy may retain registry accounts while weakening the balance sheets of the operators expected to build networks.

Bylaws article 11.3 does not close the authority inquiry either. It identifies Board ratification as a corporate step. It does not itself prove which directors validly formed the Board on 4 February, whether the required quorum acted, how conflicts were managed or what legal boundary applied while the receiver had not been publicly discharged. Nor does a corporate power necessarily settle the broader doctrine question: a private corporation’s internal documents cannot create sovereignty over non-participants merely by describing a policy mechanism.

The inadmissibility ruling is similarly limited. It may establish that the appellant did not cross a procedural threshold. It cannot be recast as evidence that the substantive objections were tested and defeated. The court posture also remains bilateral. Skyconnect’s filing does not make D3 void; AFRINIC’s survival of an unserved interim application does not validate ratification. At cutoff, the main challenge was ongoing.

Finally, a ratified label cannot substitute for running code. Resource categories must be assigned. Misclassification must be correctable. Counterpart relationships must be confirmed. Forms, tools and staff procedures must exist. The public manual must convey the operative rule. Service levels and appeal channels must be knowable. A completed transfer must demonstrate the path end to end. The public evidence did not establish all of those transitions. The exact implementation status inside AFRINIC therefore remains unknown; the public state weighs against claiming completion but does not prove that no internal work occurred.

What the numbers can—and cannot—show

The arithmetic begins with address count. For an IPv4 prefix /p, the number of addresses is:

N = 2^(32 - p)

For a /24, N = 2^(32 - 24) = 256, which is D3’s minimum transfer unit. For a /22, N = 2^(32 - 22) = 1,024. For a /16, N = 2^(32 - 16) = 65,536, equivalent in address count to 64 /22 blocks.

NRS and Heng materials cite an estimated mobility or value effect of about USD 45,000 for a /22 and USD 2.88 million for a /16. NRS appears in this record as a member-advocacy organisation: it researches, convenes and represents members who explicitly authorise it, but it does not operate AFRINIC’s registry, RPKI, WHOIS or RDAP, appeals, settlement, elections, custody or continuity. Its number is therefore an attributed advocacy scenario, not an operational registry result. The scaling is internally consistent:

USD 45,000 / 1,024 = USD 43.9453125 per address

65,536 × USD 43.9453125 = USD 2,880,000

Equivalently, 64 × USD 45,000 = USD 2,880,000. Arithmetic consistency, however, is not empirical validation. The calculation assumes the full amount per address is destroyed by the mobility restriction. The source package contains no AFRINIC-specific transaction series proving that assumption.

A 2026 broker market analysis offers lower and size-dependent reference ranges. It places early-to-mid-2026 whole /16 transactions at roughly USD 10–15 per address, giving a gross value range of 65,536 × USD 10–15 = USD 655,360–983,040. It places smaller /22-to-/24 blocks at roughly USD 28–40 per address, giving a /22 gross range of 1,024 × USD 28–40 = USD 28,672–40,960. Those are broker market references, not AFRINIC-specific verified prices. Origin, reputation, registry, demand, counterparty quality, transaction costs and block size all affect execution.

A transparent sensitivity model is:

Illustrative mobility loss L = N × P × d × q

Here N is address count, P is an assumed transferable-market price per address, d is the discount attributed to a reduced buyer or exit set, and q is a probability or realisability factor. Neither d nor q is observed in the record. Setting q = 1 only simplifies the table; it is not a forecast.

For a /22 using P = USD 28–40, gross value is USD 28,672–40,960. At a 5% assumed mobility discount, L = USD 1,434–2,048; at 15%, USD 4,301–6,144; at 30%, USD 8,602–12,288; and at 50%, USD 14,336–20,480.

For a whole /16 using P = USD 10–15, gross value is USD 655,360–983,040. At a 5% assumed discount, L = USD 32,768–49,152; at 15%, USD 98,304–147,456; at 30%, USD 196,608–294,912; and at 50%, USD 327,680–491,520.

A subdivision illustration produces a different range and execution problem. If a /16 were divided into 64 /22s and each smaller block achieved USD 30–35 per address, gross proceeds would be about USD 1.97–2.29 million. That scenario entails 64 transfers plus brokerage, fees, elapsed time, buyer search and execution risk. It cannot be compared to a whole-block sale without accounting for those differences. A leasing reference of about USD 0.38–0.50 per IP per month is another distinct category; it is not a sale price and is not automatically available for AFRINIC-origin space.

Non-fact boundary: none of these sensitivity rows is AFRINIC’s measured loss, a court award, a completed transaction price or a prediction. The NRS/Heng figures are attributed advocacy scenarios. The broker ranges are market references. The discount and realisability factors are assumptions, and small-block pricing must not be applied to a whole /16 without an explicit subdivision model.

The defensible economic claim is therefore directional, not absolute. A restriction that reduces permitted destinations can reduce optionality, narrow the legal buyer set and introduce approval uncertainty. Those mechanisms can affect bids, time to sale, financing and the incentive to bring resources into the registry. The magnitude remains unknown without prefix-level AFRINIC data and observed transactions.

The evidence demands created by the act

Because AFRINIC’s declaration moved a dormant proposal closer to a binding registry rule, the purported or acting Board should publish the evidence that distinguishes a lawful corporate act from an institutional assertion. That means the resolution, agenda, director attendance, quorum, vote, conflicts, legal basis and implementation authorisation. If some legal advice must remain privileged, AFRINIC can still disclose the authority question asked, the corporate organ that decided it and the non-privileged reasoning that connects registry necessity to origin-based exit control.

The receiver should identify the precise public-order basis, or a non-privileged account of it, under which a preservative receivership could support or permit substantive number-policy ratification before formal discharge. The public record does not establish that the receiver was discharged before 4 February. It also does not establish that the appointment order clearly authorised or clearly prohibited D3. The question is not an accusation disguised as a finding. It is the missing link between preservation, electoral transition and a structural policy act.

Policy supporters should identify the affected principals who consented to the restriction; the harm, beyond registry inaccuracy, that an outbound Regional transfer would cause; the evidence that proof of control and interoperable record exchange would not solve that harm; and the party expected to bear any liquidity effect. Invoking African development is not enough. Supporters need to show how fewer exit options produce more deployed connectivity rather than more institutional retention.

AFRINIC staff should publish the classification inventory, correction path, compatibility decisions, implementation milestones, transfer statistics and service levels. They should state where non-recognition ends and reclamation begins, how an alleged unauthorised transfer is investigated, and what independent review is available. Staff cannot repair a defect in corporate authority through technical implementation, but transparent implementation can at least prevent discretionary policy from becoming invisible operational power.

Institutional allies should separate five possible positions that are too often merged: support for technical continuity, support for holding an election, support for the receiver’s work, support for the announced Board’s corporate authority and support for D3’s substantive regional controls. Endorsing one does not entail the others. Precision is especially important when the institution calls criticism misinformation while the underlying proceedings remain unresolved.

A defensible interim position

No one needs to choose between a chaotic registry and an origin embargo. Before irreversible classification or outbound denial, AFRINIC can preserve its earlier transfer baseline while publishing the ratification and authority chain. It can run a reversible, record-only pilot that validates source control, duplicate claims, exact disputes, contact accuracy and registry acknowledgement without making origin a permanent exit restriction. It can publish error rates and processing time before expanding the rule.

The minimum defensible interim outcome is a freeze on irreversible Regional classifications, outbound denials and reclamation based solely on D3 until one of two authority routes is visible: either the court resolves the ratification challenge, or an unquestionably constituted corporate body independently reconsiders and re-ratifies the policy after disclosing quorum, conflicts and receiver boundaries. Operational readiness should then be demonstrated through a revised manual, deterministic classification rules, a correction and appeal route, counterpart mappings, service levels and end-to-end transfer evidence.

That position does not prejudge the litigation or deny AFRINIC’s declared act. It respects the difference between preserving records and converting a contested institutional moment into a durable market boundary. The question is not whether AFRINIC may keep a ledger. It is whether a ledger’s custodian may use an origin label to decide where capital can exit—and whether a Board formed through an unresolved receivership transition provided enough public evidence to exercise that power on 4 February 2026.