Summary

  • AFPUB-2026-GEN-001-DRAFT01 would replace major parts of AFRINIC’s working-group procedure, create a five-person Number Council, redesign co-chair selection, moderation, appeal and recall, and allow a petition-plus-staff route to treat policy as ratified after 60 days of Board inability or unexplained refusal. At the evidence cutoff it remained Draft 1 and Under Discussion.
  • The proposal is strongest when it specifies open discussion, chair duties, warnings, consensus records, reasons and review. It crosses the decisive boundary when a subordinate policy text reallocates the Board’s express duty to call the annual Public Policy Meeting, makes internal committee decisions unqualifiedly final, or directs staff to implement policy without the corporate act required by the operative Constitution.
  • The workable answer is neither Board supremacy nor a self-declared community sovereignty. Members must lawfully authorise a narrow continuity mechanism in the Constitution, while the policy manual supplies detailed procedure, reason-giving, conflicts, review, restoration and protection of the registry ledger.

On 20 May 2026, Grégoire EHOUMI, Noah Maina and Adeola A. P. AINA submitted AFPUB-2026-GEN-001-DRAFT01, a proposed replacement for substantial parts of AFRINIC’s Policy Development Process. It would abolish the current CPM sections 3.3 and 3.5, relocate approval, appeal and recall rules, create a five-person Number Council, prescribe how two co-chairs are selected and how they control discussion, and open a route by which a policy could be treated as ratified and implemented by staff after 60 days without an effective Board decision. AFRINIC published a staff assessment dated 21 June.

On 9 August, its proposal index still listed only Draft 1, Version 1.0, as Under Discussion. There was no evidence of consensus, Last Call, ratification, the required constitutional amendment, the associated NRO election-process change or implementation.

The dispute is not a quarrel over drafting etiquette. It concerns which actor may perform the act that changes AFRINIC’s corporate position. An open working group can debate, test proposals, document objections, select its administrative chairs and review procedural mistakes. Those are useful forms of participation. But participation is evidence, expertise and discipline; it is not a power source. A mailing list, a Public Policy Meeting, rough consensus, a Board resolution, a member vote or a Registration Service Agreement can do only the work that the relevant legal and constitutional instrument authorises.

None can enlarge its own jurisdiction by repetition or by the importance of the problem it is trying to solve.

That distinction matters because AFRINIC is a private corporate coordinator, not a regional legislature. Its proper mandatory function is thin: preserve uniqueness, accurate records, proof, notice, correction, security-relevant assertions and continuity. The registry record serves running networks; it does not create a sovereign title over them. The proposal can improve the procedure by which the working group performs its part of that coordination. It cannot, through a lower-order manual, manufacture a new corporate organ with authority that the current Constitution assigns elsewhere.

A valuable procedural constitution inside the wrong instrument

Draft 1 attempts more than one reform at once. Some elements belong naturally in a working-group rulebook. The current Consolidated Policy Manual already recognises an open Policy Development Working Group and two co-chairs. Participants discuss proposals; chairs assess consensus; a successful proposal is recommended to the Board; chair actions can be appealed; emergency variance exists. The draft does not create bottom-up participation from nothing. It writes a more elaborate operating constitution for an existing forum.

The proposed co-chair system retains two seats with staggered two-year terms. A candidate must be a natural person residing in a country within the AFRINIC service region, must be expected to participate actively and have PDP experience, and must have attended four of the six most recent Public Policy Meetings, including at least one in person. Participants present physically or remotely at a meeting would select the chairs by consensus. The remaining chair would determine that consensus; if neither chair remained, the proposed Number Council chair would do so. Tied candidates would draw lots.

If a sole candidate failed to secure consensus, the seat would be declared vacant and the Number Council would appoint an interim chair within two weeks.

Much of the proposed job description is sensible administrative craft. Co-chairs would manage process and content for fair progress, moderate and summarise list discussion, organise meetings and consultations, assign a lead chair to each proposal, assess support at the meeting and consensus after Last Call, and recommend successful text to the Board. Writing these duties down makes informal power visible. It gives participants a vocabulary with which to challenge delay, silence or arbitrary improvisation.

The detail also reveals the reach of the office. Co-chairs could reject or defer contributions categorised as old, minor, mistimed or outside scope. They could restrict speaking at a meeting if the Working Group accepted the restriction. Mailing-list discipline would normally escalate from direct communication to at least one public warning and then, as a last resort, a posting suspension of no more than 30 days. A further restriction could be longer if conduct resumed or worsened, but Draft 1 supplies no ultimate ceiling. A suspended participant would continue receiving messages, and a suspension would be appealable.

This is the kind of internal platform moderation a working group needs. It is not registry punishment. Losing posting privileges is different in kind from losing a number resource, a registry record or a security service. The distinction must remain hard. A chair can keep a discussion usable; neither a chair nor a policy room thereby gains authority over assets or staff beyond its delegated procedural remit.

Draft 1’s omissions matter even at this internal level. It sets no express evidentiary standard for suspension, no deadline for an appeal, no automatic stay while that appeal is heard and no ceiling for a repeat restriction. It relies on consensus for policy and office selection without defining consensus, a gap AFRINIC’s Policy Liaison expressly identified. Its attendance screen uses four of six meetings, including one physical attendance, as a proxy for experience. That ratio is two-thirds, but the in-person requirement also converts travel budgets, visas, health, caregiving and meeting continuity into eligibility for procedural office.

Attendance can show exposure to a process. It cannot prove a mandate from operators who were absent.

The five-seat junction

The proposal’s institutional centre is a new Number Council. Its five members would be the two PDWG co-chairs, the two community-elected NRO Number Council or ASO Address Council representatives, and the one Board-appointed representative. One of the two community-elected NRO members would chair it. Decisions would be made by a majority of members present.

The Council would publish qualified co-chair candidates, appoint interim co-chairs in specified vacancy cases, administer parts of chair and NRO continuity and call a Public Policy Meeting at least once a year. AFRINIC’s staff legal assessment did not treat every possible Council role as forbidden. It found no categorical impediment to a body that facilitates chair selection. That is an important concession: the corporate boundary does not require the Board to micromanage the Working Group.

Yet Draft 1 gives the Council no express quorum. “A majority of members present” says how a decision is counted after attendance has been established; it does not say how many of the five must attend before the body can act. It gives no general recusal rule, no complete replacement mechanism for every vacancy and no review path for appointment actions described as final and binding. The structure also overlaps the very offices it supervises. Both co-chairs sit on the Council. All three of AFRINIC’s NRO representatives sit there too. Institutional memory is concentrated, but so are appointment, meeting and review connections.

The clearest conflict is the annual meeting. The operative Constitution’s article 11.2 assigns the Board the duty to call a Public Policy Meeting at least once each year. Proposed section 3.3.8 assigns the annual call to the Number Council. The two provisions use the same minimum and name different actors. Draft 1 itself acknowledges that article 11.2 must be amended. It also says the NRO Number Council election process must be amended to fit the new design.

That admission is not a footnote. It identifies the proposal’s constitutional dependency. Until members validly amend the Constitution, a CPM clause cannot displace the Board’s express duty. A rulebook can describe how a meeting is chaired, who may speak, how support is tested and what record must be published. Calling the meeting in place of the constitutionally named organ is a different act.

The difference becomes sharper under Mauritius company law. Companies Act section 129 places the company’s business and affairs under the management, direction or supervision of the Board, but makes that authority subject to the Act and the company’s constitution. Section 143 requires directors to act within those limits, in good faith, for proper purposes and in the company’s best interests. The Board therefore has real authority and real limits. It is not a rubber stamp, and it is not a sovereign above the Constitution.

The correct reading is interdependence: statute, Constitution, members, Board, working group and staff each act within a defined layer.

The Constitution supplies the route for changing that allocation. Registered and Resource Members, acting jointly through the stated corporate process, may amend, replace or revoke it by special resolution. The threshold is 75 per cent of votes cast by entitled members voting on the question. An open PDWG consensus is not that corporate act. Nor can the corporate vote, by itself, turn AFRINIC into a public sovereign. It can amend the company’s own Constitution within law. The scope of every act remains bounded.

Appeal is not one thing

Draft 1 also rebuilds appeal and recall around a single three-person panel: the two community-elected NRO representatives and the immediate past co-chair. The same body would serve as both Appeal Committee and Recall Committee. That economy may be attractive in a volunteer system, but it joins functions that demand different kinds of independence.

For a posting suspension, the panel would hear the co-chairs and complainant and could annul a restriction it found disproportionate or unjustified. For a broader challenge to a co-chair action, the appellant must first seek resolution with the chairs or PDWG, secure support from three people who took part in the discussion and file within two weeks of public knowledge. The committee could annul a decision if the consensus determination or PDP had not been followed. For a requested recall, one person could file a justified request with at least ten other supporters from ten different organisations.

Each supporter would need at least one year of mailing-list subscription and attendance at one meeting in the previous two years. The panel would hear the parties, decide within four weeks and publish a report. Its stated grounds would be inability, deliberate obstruction of progress, or constant and persistent breach of the PDP or Working Group rules.

Automatic recall is more abrupt. A co-chair would be removed after missing two consecutive meetings without reasons, or after the other co-chair publicly reported that the person was no longer attending to Working Group affairs without reasons. Draft 1 states no independent confirmation, notice, proof or response step for the second route. Across these mechanisms, the panel’s decisions—and the Number Council’s chair-appointment actions—are described as final and binding.

Finality can end procedural limbo. But it cannot enlarge jurisdiction. The draft does not say whether “final and binding” binds only the Working Group, also binds the Board and members, or purports to exclude ordinary courts. A private manual cannot exclude statutory or judicial authority merely by using conclusive words. AFRINIC’s staff assessment accordingly questioned unqualified finality and blurred committee jurisdiction.

It also identified the missing conflict architecture: the immediate past chair may be unavailable; an NRO representative may be reviewing a close colleague; the same small panel may hear both an appeal about process and a recall of the person who ran it.

The Registration Service Agreement contains another appeal, but it is not the missing appellate tier here. RSA clause 13 concerns an organisation or member appealing from an assigning registry to the parent registry and ends in a final AFRINIC Board decision. It does not review a PDWG co-chair’s moderation, consensus or procedural act. Calling both mechanisms “appeal” does not make them interchangeable. The object, standing, decision-maker and remedy differ.

The 60-day transformation

The draft’s most consequential clause begins with an ordinary path. Co-chairs would send a consensus proposal and the discussion and Last Call report to the Board, which the draft says shall ratify. That is already stronger than a mere recommendation, and it raises the question of what lawful grounds remain for Board review. The proposal then adds a bypass.

If the Board is unable to act, or refuses to ratify without giving reasons, for 60 days, any PDWG member could start a petition. It would require support from at least 15 PDWG members who are registered contacts of 15 distinct AFRINIC members and who collectively come from at least three AFRINIC subregions. AFRINIC staff would perform a new analysis. If staff identified no new issue, the policy “must be considered ratified” and implemented by staff.

This mechanism tries to solve a genuine failure: a working group should not spend months developing a proposal only for silence or a non-quorate Board to freeze it indefinitely. Requiring reasons and a deadline is a useful discipline. Distribution across distinct members and subregions is better than allowing a single faction or one company’s contacts to trigger the result. A fresh technical and legal check is better than implementation without examination.

But the terminal act has changed. Under the operative structure, a working-group recommendation becomes policy through the constitutional approval chain. Under the bypass, a petition threshold plus the absence of a “new issue” in staff analysis produces deemed ratification and directs implementation. Staff has moved from adviser and implementer to the last condition that makes the policy effective. The draft may preserve the word ratification; it changes who performs its substance.

Several definitions that could constrain this power are missing. “Unable to act” could mean no quorum, vacancies, a legal restraint, delayed scheduling or practical paralysis. Draft 1 does not distinguish them. It does not state when the 60-day clock starts, whether a partial explanation stops it, what counts as reasons, who verifies the 15 contacts, whether the initiator counts among them, which subregion taxonomy applies or when contact authority is measured. A database contact is not automatically authorised to perform every legal act for a member.

The clause also does not define a “new issue”, materiality, cure or an appeal from staff’s conclusion.

Most importantly, the proposal acknowledges that the Constitution needs amendment but includes no explicit commencement sequence. It does not say that the new PPM and bypass provisions remain dormant until a valid member resolution has taken effect and the NRO election rules have been aligned. A continuity device that operates before its authority exists risks producing exactly the uncertainty it was written to remove.

Why a procedural defect reaches operators

Draft 1 changes no allocation, transfer, routing or security rule by itself. AFRINIC’s staff assessment marked no impact on WHOIS, RDAP, MyAFRINIC, NetSuite, NMRP or RPKI—six named systems—and stated no Member Services, IT or human-resources impact. It noted costs for community consultations but did not quantify them. Those findings are relevant and should restrain exaggeration. This is not a report of an outage or a completed adverse resource decision.

Zero immediate system change, however, is not zero economic effect. Procedure selects which future policy can advance, whose objection enters the consensus record, who may reverse a chair, why a Board may withhold approval and when staff may implement. The RSA defines policy with downstream contractual significance by reference to due adoption under the Constitution. Once a validly adopted resource policy applies through services and member obligations, the quality of the authority chain is no longer internal housekeeping.

An invalid or ambiguous chain creates a delayed risk. Operators may plan, contract or request registry action under a policy that later becomes disputed. AFRINIC may have to defend the validity of implementation. Members and counterparties may price uncertainty into transactions. The published RSA compounds the concern by capping liability at the greater of six months of fees or US$100. The closer the process comes to decisions capable of changing operational and economic conditions, the less defensible it is to pair concentrated discretion with a near-symbolic downside.

This is why the registry’s role must remain modest even while its procedures become better. Accurate records, proof, notice, correction and continuity create value because operators can rely on a predictable ledger. Punitive, legislative or sovereign pretensions create a governance-risk premium because the actor controlling the chokepoint need not bear the full cost of error. Detailed procedure is valuable when it limits that risk. It is dangerous when detail becomes a substitute for authority.

The strongest case for the draft

The case for a continuity backstop should be stated in its strongest form. AFRINIC’s working-group process had been described in October 2025 as effectively paused during years of limited corporate functionality. Depending on a functioning Board for every meeting, appointment and policy conclusion leaves a single institutional point of failure. Sparse current rules have also contributed to recurring disputes over chair selection, consensus, moderation, appeal and recall. A working group that cannot renew its own administrators or obtain reasons for a stalled recommendation is not meaningfully bottom-up.

Draft 1 responds with visible machinery. It spreads roles among co-chairs, elected NRO representatives, a Board-appointed representative and staff. It makes warnings precede suspension, preserves receive access, creates an appeal, publishes recall reports, imposes a four-week recall deadline and makes Board silence contestable after 60 days. Its authors defended the integrated design as a self-sustaining mechanism for a community-driven PDWG. Supporters on the RPD list treated continuity during Board failure as a bottom-up necessity that should exist alongside broader constitutional repair.

That is a serious case. The answer cannot be to restore unlimited Board discretion or to make working-group operation hostage to one missed quorum. AFRINIC’s own legal assessment did not propose that. It said a refusal to ratify should rest on reasonable, objective and clearly articulated grounds.

Its suggested residual intervention power would require good faith, necessity and proportionality; would confine action to legality, the company’s protection, operational continuity, material governance failure and significant legal, regulatory, financial or operational harm; would preserve substantive policy independence as far as practicable; and would require consultation with chairs or the Council where practicable.

Those constraints reveal the possible settlement. The Board remains responsible for the corporate act, but it cannot use that role as an unexplained substantive veto. The Working Group remains the place for resource-policy judgment, but it cannot transform participation into corporate sovereignty. A narrow, reviewable boundary serves both continuity and restraint.

RFC 2418 supports detailed working-group administration: clear chair craft, open discussion, moderation and an organised route through disagreement. It also preserves an external IETF oversight chain. It is a procedural comparator, not a source of Mauritian corporate authority. Borrowing its useful techniques cannot answer who may bind AFRINIC or direct AFRINIC staff.

The NRO process supplies another bounded context. On 28 July 2026, the ASO Address Council delivered version 3 of a recommended RIR Governance Document to the NRO Executive Council. Review, coordination and the route toward a final document remained pending. At the cutoff it was a recommended draft, not adopted authority. Global recognition criteria may define future failure and continuity expectations. They do not amend AFRINIC’s domestic Constitution.

A correction that shows why records matter

Draft 1’s problem statement said the RPD mailing list had been frozen. Ben Roberts promptly disputed that description. On 9 June, Grégoire EHOUMI acknowledged the error in an authors’ digest and said Draft 2 would describe the list as inactive. This distinction is not semantic housekeeping. A frozen list suggests deliberate prevention of participation; an inactive list describes a channel that was not functioning in practice without proving why.

No Draft 2 appeared on AFRINIC’s proposal index by the cutoff. The record therefore has two parts that must both be preserved: Draft 1 contains the incorrect wording, and an author publicly accepted the correction and promised a future revision. The corrected phrase cannot be silently written back into Draft 1, just as the error cannot be repeated as fact after its acknowledgement.

The same discipline applies to the rest of the evidence. RPD messages establish what named participants said, not a vote or institutional finding. Sami Salih, a former co-chair, supported the continuity objective but urged that governance, electoral, disciplinary, operational and procedural questions be separated. Abdulkarim Oloyede questioned physical-attendance eligibility, subjective consensus selection and final recall. Seun Ojedeji opposed the current draft while supporting a Board reasons requirement.

Andrew Alston, writing personally, distinguished operational rules from governance rules and proposed reasoned Board review rather than silence or automatic substitution. Authors and supporters defended PDWG self-sufficiency. The archive proves the disagreement; the operative instruments decide the authority available at the cutoff.

AFRINIC staff’s legal assessment is equally bounded. It is the institution’s legal view, not a Mauritian court holding. It carries weight because it compares the proposal with AFRINIC’s own constitutional structure and recommends text, but a court could face facts or arguments not contained in the assessment. Institutional analysis must distinguish a persuasive warning from an adjudicated result.

Two layers, one lawful sequence

The sound counterfactual begins with a member-approved constitutional amendment. It should define what counts as Board incapacity, who certifies it, who may temporarily call a Public Policy Meeting and how the temporary authority ends. It may charter a Number Council, but the charter must set quorum, vacancies, recusals, reasons, minutes, terms and review boundaries. It must state whether, and under what narrow conditions, a ratification backstop exists. Staff should never become the accidental ratifier merely because it reports no new issue.

The policy manual should then do the work it is suited to do. It should specify chair duties, open participation, consensus records, moderation stages, suspension ceilings, expedited appeal, recall evidence, hearing rights, conflicts and published reasons. A Board should approve or issue a reasoned objection within a fixed period and only on listed legal, constitutional, fiduciary or implementability grounds. An objection should return the proposal to the Working Group with a cure path. It should not become a silent veto or a licence for the Board to substitute its preferred resource policy.

If the Board is legally incapacitated, a constitutionally authorised temporary mechanism can preserve the last valid state, permit the meeting and obtain independent legal and process certification before staff acts. Emergency authority must sunset when ordinary quorum returns. It must include a restoration rule, so a temporary continuity bridge does not harden into a permanent rival government. It must preserve the verified consensus text rather than creating authority to rewrite substantive policy during the emergency.

This arrangement refuses a false choice. The ledger does not need the permanent supremacy of a gatekeeper, but continuity of its accurate state does matter. The Working Group does not need a sovereign mandate to perform valuable technical and procedural work, but the corporate act that activates policy must have a lawful source. The Board does not need unlimited discretion to discharge its legal responsibility, but it must give reasons and remain reviewable. Members do not become regional legislators by amending the Constitution, but they can lawfully define the company’s internal delegation and failover.

Draft 1 contains the ingredients of a better working-group system and the diagnosis of a real institutional weakness. Its mistake is not that it wants continuity. Its mistake is asking a subordinate manual to supply authority before the constitution it depends on has changed. The remedy is to put each rule in the instrument capable of carrying it, in the right order, with power matched by reasons, review, liability and restoration.