Summary

  • AFRINIC's 12 March 2026 member update said that Gowtamsingh Dabee remained the court-appointed receiver pending formal discharge, that judgment on his discharge application was awaited, and that he had been approving IP address allocations and assignments which AFRINIC described as having properly passed through the application process.
  • That sentence identifies the claimed approver; it does not publish the operative court-order clause, a lawful delegation, the limits of the approval power, or the request-level records needed to establish who decided what and why.
  • The strongest case for continuing allocations is substantial: legitimate operators, universities, Internet Exchange Points and other applicants should not be stranded by institutional litigation, and a receiver may hold broad operating powers. But urgency and service continuity make a bounded, reviewable approval protocol more necessary, not less.
  • AFRINIC is a technical bookkeeper and ledger coordinator, not a sovereign or punishment authority. It should preserve an accurate, non-retaliatory registry while publishing a chain from legal authority through staff review, final decision, payment and agreement, registry execution, correction and handback.

A sentence with the weight of a signature

Institutional admissions are sometimes disguised by ordinary grammar. The consequential line in AFRINIC's member update of 12 March 2026 did not arrive as a court order, a schedule of delegated powers or a register of decisions. It arrived as an explanation. The people AFRINIC described as its Board said that Gowtamsingh Dabee, the court-appointed receiver, had been approving IP address allocations and assignments that, in AFRINIC's description, had properly gone through the application process. The same update said that his application for discharge had been heard, that judgment was awaited and that formal discharge remained pending.

Read narrowly, the line supplies useful information. It names the person whom the purported Board said occupied the approval point. It places that conduct during an unresolved receivership rather than after a completed handback. It also tells members that AFRINIC regarded the applications as properly processed. Those are material public statements by the institution operating the regional number registry.

Read carelessly, however, the line can do work it is incapable of doing. It can make an identified actor look like an authorised actor; make completed processing look like proof of lawful approval; and make a functioning database look like ratification of everything that preceded the entry. None of those propositions follows from the disclosure itself. A statement about who approved is evidence of claimed operational control. It is not the instrument that confers that control.

That distinction matters because the same update acknowledged a live dispute. AFRINIC attributed to Cloud Innovation Ltd the contention that the receiver had acted outside his authority by enabling the provision of number resources to new and existing network operators, academic institutions, Internet Exchange Points and corporate organisations in the service region. AFRINIC also said that a separate application sought an injunction preventing it from allocating IP resources and that it was resisting the application. These are attributed litigation positions, not judicial findings.

The available public record does not establish that the contention was correct, that the injunction was granted or refused, or that a particular allocation was valid or invalid.

The responsible conclusion is therefore neither absolution nor accusation. The update proves that the purported Board said the undischarged receiver was approving. It does not prove the legal source, scope or exercise of that authority. Nor does the absence of the operative material on the checked public record prove that no private, unindexed, protected or later document exists. What it does establish is a public accountability deficit proportionate to the significance of the admission.

The chronology makes the missing chain more important

Dabee was appointed receiver on 12 February 2025 in place of the Official Receiver in a proceeding before the Bankruptcy Division of the Supreme Court of Mauritius. AFRINIC's appointment notice described the public purpose as securing the company's assets and seeing its Board reconstituted under the bylaws. A communiqué issued in April 2025 described the receiver's role in terms of holding the ring, preserving the status quo and business value, and reconstituting the Board. These descriptions explain why operational preservation mattered.

They do not reproduce the authenticated appointment order or every later variation, and they do not identify a specific clause governing number-resource approvals.

In March 2025, ICANN told Dabee that it understood assignments had been limited under the Official Receiver. It cited at least one applicant complaint and urged a restart under Dabee's direction while the election process continued. That intervention illuminates a real continuity problem. Applicants cannot build networks from institutional assurances alone; delays affect deployment, customers and investment. Yet ICANN coordinates identifiers across systems. It does not confer Mauritian judicial or corporate power. Its request could support the operational case for action without becoming the legal source of that action.

A receiver-signed communiqué dated 22 July 2025 said that new IPv4 and IPv6 allocations had resumed exceptionally on 1 July to clear a backlog. This was not a momentary response at the edge of the March 2026 disclosure. The restart preceded it by 254 calendar days. The receiver's appointment preceded it by 393 calendar days. And the receiver had applied on 8 October 2025 for termination of the receivership and formal release, 155 calendar days before the March update, yet the update still described the discharge judgment as awaited and the receiver as the person approving resources.

Elapsed time does not create or extinguish a legal power. It does change the scale of the accountability question. A short emergency intervention might concern a handful of reversible acts. An approval function maintained across months can create a portfolio of recipient expectations, signed agreements, payment records and registry objects. Each additional decision may be perfectly supportable. Each also increases the cost of failing to preserve the instrument-to-transaction chain when the decision is made.

Later operational figures show why output cannot substitute for provenance. IANA recorded an upstream allocation of 1,024 ASNs to AFRINIC on 25 February 2026, covering AS329728 through AS330751. IANA's figures based on AFRINIC data also reported 88 ASN allocations to network operators across February to July 2026: monthly counts of 10, 10, 23, 16, 16 and 13. These numbers demonstrate observable registry activity. They do not show that Dabee personally approved any listed ASN, and they do not establish whether the purported Board's wording about IP address allocations and assignments was intended to include ASN decisions.

An upstream inventory allocation to AFRINIC is also different from a downstream allocation to an operator. Conflating the two would turn continuity data into false evidence about an internal approval.

Approval is not the whole process

The apparent simplicity of the word “approving” conceals a chain of decisions. A resource request first has to enter an identified queue. The applicant's identity and eligibility have to be checked. Technical staff examine evidence of need, utilisation or other criteria under the applicable policy and determine the appropriate resource type and size. Finance may issue an invoice. The applicant and AFRINIC execute the relevant Registration Service Agreement. Authorised staff register the resource, update WHOIS or RDAP and perform related operational work.

Depending on the resource and use, later steps may involve reverse DNS, routing registry information or other security-adjacent records.

At least seven control stages are readily distinguishable: request intake; identity and eligibility review; need and size review; approval; invoice and payment; agreement execution; and registration. Routing-support work follows. The stages need not carry these exact internal names in every period or for every resource. Their separation nevertheless exposes the analytical error in treating a finished registry record as a single act by a single institution.

The March sentence does not say whether Dabee personally assessed technical need, accepted a staff recommendation, approved only certain categories, signed a batch, delegated routine decisions subject to thresholds, or merely supplied a final authorisation after operational staff had completed their work. It does not say which policy version governed a request, what inventory was available, who checked the applicant, who calculated the resource size, who resolved an exception, who held signing authority for the agreement, or who used the credentials that changed the registry.

It also does not say whether the words “allocation” and “assignment” were being used with the precision found in the policy manual.

That precision is essential. An allocation to a Local Internet Registry for further distribution is not the same as an assignment for use by an end site or network. A downstream assignment is not a sub-allocation, a transfer, an upstream delegation from IANA, or the later act of registration. Policy can set criteria for distributing a genuinely unallocated free pool. That legitimate function does not create a permanent ownership interest over resources once they are operational, and it does not give a registry jurisdiction over an operator's wider business merely because the registry once recorded the resource.

The practical file for a defensible request should therefore be boringly specific. It should show the date and queue position; the applicant and eligibility evidence; the policy version; the technical calculation; the inventory snapshot; the staff recommendation; the identified final approver; the legal instrument or delegation supporting that approver; the decision and reasons; the conflict check; the invoice and payment reconciliation; the agreement version and lawful signatory; the registry change and its timestamp; and the route for reconsideration, correction or appeal. Sensitive applicant information need not be published.

The existence, integrity and accountable custody of the file must be demonstrable.

Without that file, several different kinds of legitimacy are allowed to blur together. Technical legitimacy asks whether the request met the substantive policy. Corporate legitimacy asks whether the company actor could make or delegate the decision. Judicial authority asks what the receiver was empowered to do under the exact appointment and variation orders. Contractual legitimacy asks whether the agreement was executed by authorised parties. Operational legitimacy asks whether the registry change faithfully implemented the decision. Success at one stage cannot silently repair a failure at another.

What the public material does and does not establish

The public appointment notice establishes the date and stated purpose of Dabee's appointment. The Insolvency Act supplies a framework under which a receiver can have broad powers to operate a company, subject to the appointing instrument and court orders. The Companies Act, AFRINIC's bylaws and its policy manual identify relevant corporate and policy surfaces. The receiver's July communiqué establishes that he publicly announced a restart of new IPv4 and IPv6 allocations. The March update establishes what the purported Board said about his approval role. Together, these materials make a broad-power explanation plausible.

They do not close the transaction question. The checked public record did not disclose the authenticated full appointment order and all variations affecting resource decisions; a dated receiver instruction or delegation covering allocations and assignments; a court-approved continuity protocol; or a responsibility matrix separating the receiver, purported Board, Registration Services, Hostmaster, Finance and signatories. Nor did it disclose a complete decision log containing the application, policy analysis, staff recommendation, final decision, reasons, agreement, payment, registry timestamp, conflicts and review outcome for each request.

There are also unresolved endpoints. The public material checked through 10 August 2026 did not establish the final disposition of the application seeking to restrain allocations or a final order discharging the receiver. It did not establish the current lawful scope of the purported Board's oversight while receivership remained formally unresolved. These are bounded public-record findings, not declarations that the documents do not exist anywhere. They are reasons to demand publication or controlled independent access, not invitations to invent a verdict.

This evidential discipline cuts both ways. AFRINIC's statement that applications properly passed through its process is an institutional assertion, not independent proof of compliance for every file. Cloud Innovation's outside-authority argument, as described by AFRINIC, is a litigant position, not a judgment. An injunction application is not an injunction. A discharge application is not a discharge. A recipient's appearance in a registry is not proof of the approver's power.

Equally, missing public detail is not proof that an applicant acted improperly, that a staff member exceeded instructions, or that the receiver lacked power in law. No such individual finding follows from the available evidence.

The case for continuity deserves its strongest form

The strongest defence of the restart begins with people and networks rather than institutional prestige. A university waiting for address space, an Internet Exchange Point expanding connectivity or an operator deploying service should not bear the cost of a corporate dispute it did not create. Scarce free-pool resources affect network planning, capital spending, customer acquisition and service availability. A frozen queue can become a quiet form of economic damage.

A receiver charged with preserving AFRINIC's business may need to continue ordinary operations. Mauritian insolvency law can provide broad operating powers, although their application depends on the appointment instrument and relevant court orders. Objective review by trained staff under an established policy is safer than improvised political distribution. The announced July 2025 restart addressed a backlog. Requiring a fresh court hearing or a purported Board resolution for every routine technical request could make the registry unusable and defeat the continuity the receivership was intended to protect.

That is not a weak argument to be brushed aside with formalism. It correctly identifies the harm of paralysis and the possibility that a receiver's authority was wide enough to cover routine approvals. It also recognises that continuity need not mean that the purported Board personally decides each application. A functioning approval service can be delegated within strict rules.

But the defence points towards documentation rather than deference. If approvals were routine, a standing direction could define resource types, thresholds, staff roles, exceptions and a sunset date. If the order granted broad powers, the operative clause could be identified and, subject to necessary redactions, published. If staff applied objective criteria, their recommendations and calculations could be preserved. If decisions were numerous, standardised maker-checker controls and immutable logs would reduce the cost of proof. Routine acts are easier, not harder, to document consistently.

Urgency may justify a narrow and reversible continuity protocol. It cannot justify permanent authority inflation. ICANN's encouragement cannot become a substitute for a Mauritian instrument. Applicant need cannot create powers in the actor processing the application. Policy compliance cannot cure an absent corporate decision-maker. And the absence of an outage cannot retroactively prove that the approval chain was lawful. These propositions protect recipients as much as they constrain institutions: a good-faith applicant benefits when its approval can survive a later challenge without depending on the prestige of whoever was in charge.

A bookkeeper must not audition for sovereignty

AFRINIC performs work that is technically central and institutionally narrow. It coordinates a regional ledger so that number resources are unique, records are accurate, requests are processed, security metadata is maintained and operational continuity is possible. Centrality makes good records important. It does not turn a Mauritius private company into a sovereign, legislature, regulator, court, prosecutor or owner of the resources used by operators.

The same boundary applies to everyone acting through AFRINIC. A receiver's title does not manufacture unlimited power. A purported Board's label does not validate individual approvals. A staff recommendation does not supply judicial authority. An external organisation's recognition does not decide Mauritian corporate law. A policy community cannot create punishment powers by calling its preferences stewardship. Authority belongs to the exact legal, corporate and contractual instruments that govern the act at the time it is taken.

This matters beyond the initial allocation. A registry controls practical chokepoints: request intake, database records, agreement administration, reverse DNS and routing-adjacent information. Those tools must not be used to punish litigation, criticism or a disfavoured business model unless an exact lawful power, applied to the particular matter and subject to independent review, permits the action. An unrelated dispute must not contaminate an operator's routine services or its customers' assignments. The registry's duty is to keep a faithful ledger, not to convert a record-keeping dependency into political dominion.

The free pool illustrates the limit. When resources are genuinely unallocated, documented need and eligibility criteria can govern scarce distribution. The bookkeeper may apply those criteria because orderly allocation requires a neutral rule. Once resources are operational, that initial administrative role does not mature into continuing ownership of the resource or control over the recipient's whole enterprise. The policy mirror must reflect the narrow function back to the institution: criteria for distribution are not a deed of sovereignty.

Continuity should therefore protect the ledger, not the gatekeeper. The protected objects are the accurate queue, the last verified registry state, valid applicant files, security information, credentials under controlled custody and the ability to correct errors. No receiver or purported Board is entitled to institutional immunity simply because these functions must continue. The better the continuity architecture, the less the system needs to rely on the personality or asserted status of any one actor.

The immediate cost falls on recipients and members

An opaque authority chain is often described as a governance problem, as if its cost were confined to meetings and legal argument. The more direct exposure belongs to recipients. An applicant may have supplied every document, satisfied every technical criterion, paid every invoice and signed every agreement in good faith. If the institution cannot demonstrate who lawfully gave final approval, that recipient inherits a provenance risk it did not create.

The risk need not result in reversal to have an economic effect. A later court, lawful successor, member or counterparty may ask whether the decision was within power. The recipient may face delay, uncertainty, additional diligence or a discount in financing and commercial negotiations. Staff may be unable to explain the chain without reconstructing it from email and database traces. A future decision-maker may confront a false choice between treating all past actions as unquestionable and destabilising innocent networks. Proper contemporaneous records prevent that trap.

Members face a related exposure. Their fees support the service, and the registry's conduct affects the reliability of the common ledger. Yet member silence, continued use of services or later approval of accounts cannot create an authority that was absent when an individual decision was taken. Aggregate reporting should let members see queue size, approval and rejection counts, processing times, inventory movements and exceptions without disclosing sensitive applicant evidence. That information allows oversight of distribution while respecting privacy.

Staff require protection too. A hostmaster or finance officer should not have to infer the boundary between a court-appointed receiver and a purported Board from competing public statements. A clear instruction should identify who can decide, who can sign, who can execute, which matters are reserved and where staff can escalate uncertainty. The file should distinguish substantive technical judgment from the legal authority to accept that judgment. Staff processing is evidence that work occurred; it is not a substitute for the mandate behind the final decision.

The governance dispute should also be isolated from legitimate applications. A blanket halt would make applicants collateral damage. An undocumented restart creates the opposite risk by allowing uncertainty to accumulate in completed decisions. The answer is documented lawful continuity: keep ordinary bookkeeper services running, protect the last verified state, preserve a route to correction and prevent contested institutional power from becoming a weapon against operators.

The records that would answer the admission

The most useful response to the 12 March sentence would not be another communiqué. It would be an authority ledger built from instruments and transaction records.

First, publish the operative appointment order and every variation that affects resource decisions, with only legally necessary redactions. Identify the clauses said to authorise the receiver to operate the allocation service. If a court direction or later order altered that power, show its effective dates. If confidentiality prevents full publication, give controlled access to an independent reviewer and publish a sufficiently precise account of scope.

Second, disclose the receiver's dated direction or delegation. It should state which resource types were covered, whether approval was personal or delegated, what thresholds applied, who could handle exceptions, what matters were reserved and when the instruction expired. “Carry on business” may be a legal conclusion; staff need an operational rule.

Third, preserve a request-level chain. Each file should connect its queue entry to the applicant evidence, policy version, technical calculation, inventory position, recommendation, approver, reason, conflict declaration, invoice, payment, agreement, signatory authority, registry change and review result. The public does not need the applicant's confidential network plans. A court, reviewer or lawful successor must be able to establish that the chain exists and has not been reconstructed after the event.

Fourth, establish independent review and correction. A disputed approval should have a filing window, an identified reviewer, access to reasons, interim protection against retaliation and a recorded outcome. Registry corrections should preserve before-and-after states. A challenge to authority should not itself cause an operational resource to disappear before an independent decision.

Fifth, report the portfolio in aggregate. Queue age, processing time, approval and rejection totals, resource type, inventory consumption, exception rate and review outcomes would show whether continuity is serving applicants fairly. These figures are more informative than repeated claims that the process was proper because they reveal whether similar requests were treated similarly.

Finally, prepare a signed handback. On formal discharge, the lawful successor should receive the request queue, decision files, unresolved reviews, inventory reconciliation, agreement register, credential custody and a record of every open exception. A receivership should end with a reproducible ledger, not a pile of institutional assumptions.

The 12 March update made the approval question public. The answer must now be public enough to be tested. Treat the sentence as an admission of claimed operational control, not proof that the control was valid. Continue processing legitimate requests, but make every approval traceable to a lawful power, a reasoned file and an independent route of correction. That is how a registry remains useful without pretending to be a sovereign.