Summary

  • On 28 April 2018, AFRINIC published PDP Bis Draft 3 as a proposal under discussion, not an adopted or implemented policy. At AFRINIC-28 in Dakar on 9 May, the recorded outcome was “More discussion needed”.
  • The draft improved procedural visibility but placed decisive gates with two chairs, counted a silent Last Call as consensus, combined a narrow Board ratification review with a separate urgency route, and offered an appeal through three institutionally connected seats.
  • Those mechanisms could administer AFRINIC’s private registry process. They could not turn attendance, mailing-list activity, non-response, fiduciary language or an internally “final” decision into sovereign authority or authorization from affected principals.
  • A durable replacement would confine mandatory policy to the registry’s technical necessities, make consent evidence positive and auditable, bound Board review to implementability and corporate legality, and separate private process review from the people whose decisions it examines.

L3 — In Dakar, a quiet room could still become consensus

On 9 May 2018 in Dakar, two co-chairs sat at the front of an unusually consequential procedural question. The text before the meeting would allow a policy proposal to pass through a Last Call in which no feedback counted as consensus. It would then travel to a Board instructed to ratify unless one of two stated grounds justified remand, while a separate clause described an urgent route by which the Board could adopt number-resource policy before later Working Group consideration. A participant wishing to challenge a chair action or consensus decision would need three participating supporters and would have two weeks to file.

This was not merely meeting etiquette. It was a proposed constitution for decision-making inside a private registry.

That description needs an immediate boundary. AFRINIC is a private, membership-based technical bookkeeper and coordinator. It maintains number-resource records and supports the practical conditions under which unique identifiers can be administered coherently. It is not a government, legislature, regulator, police force, prosecutor, punishment authority, confiscator or court. The importance of its records does not transform its internal policy procedure into public law. The question in Dakar was therefore not whether procedure matters. It plainly does.

The question was whether the proposed procedure kept its powers within the registry’s narrow technical and contractual function, and whether the evidence used at each gate was strong enough for the consequence assigned to it.

The exact instrument was AFPUB-2017-GEN-002-DRAFT-03, Version 3.0. It had been submitted on 28 April 2018, was marked “Under Discussion”, and proposed to replace section 3.0 of the Consolidated Policy Manual. That status matters twice over. First, Draft 3 was a proposal rather than settled policy. Secondly, the Dakar outcome did not change that fact: the official meeting record says that PDP Bis Draft 3 required more discussion. The available record does not show Draft 3 being adopted, ratified or implemented. Nor does it show its appeal machinery being used or its emergency route being exercised.

Institutional analysis must not promote a proposed pathway into an accomplished act.

The revision history identified five changes for Draft 3 itself. It clarified that “consensus” in the policy development process meant rough consensus. It clarified how a document editor would be selected. It clarified Board ratification or non-ratification. It clarified Board variance and removed section-specific references to the bylaws. And it moved the start of the implementation clock from consensus at Last Call to Board ratification. These were meaningful corrections. They addressed ambiguity at the points where text is edited, agreement is declared, corporate review occurs and operational work begins.

Not every feature attached to Draft 3 was new on 28 April, however. The recorded changes for Draft 2 had already included implementation provisions, removal of restrictions on Board variance, appeal deadlines, amendments to the appeal route, separation of the policy process from the Working Group guidelines, and a move to two chairs. The attached Working Group guideline was dated 9 November 2017 and had been submitted with Draft 2. Draft 3 inherited that architecture and incorporated the guideline into the proposed arrangement.

Without a complete line-by-line redline, it would be wrong to describe every two-chair, moderation or appeal clause as a fresh Draft 3 invention. The proper subject is the power map that Draft 3 presented in April 2018, while distinguishing its five recorded revisions from inherited machinery.

The proposal answered genuine complaints about procedural incompleteness. Its problem statement identified missing controls at the adoption stage, an undefined idea of consensus and the absence of a policy-process provision for Board adoption under the constitutional route it cited. A process that moves consequential proposals without visible stages, definitions or reasons invites avoidable uncertainty. Draft 3’s answer was to divide the route into Adoption, Discussion, Review and Concluding phases. The names made the lifecycle easier to follow.

A proposal could be screened, debated, refined, tested and brought to closure rather than drifting through an unlabelled exchange.

The chairs were the connective tissue of those phases. They screened problem statements for clarity, scope and duplication. They could dissuade proposals judged outside scope or duplicative. They managed summaries, editing and movement from one stage to the next. They determined whether consensus existed after considering both meeting and mailing-list discussion. Once consensus had been declared, the co-chairs were to send the Board a comprehensive report covering the proposal’s lifecycle.

These duties could prevent an immature submission from consuming the group indefinitely, preserve a coherent record and ensure that the Board saw more than a bare final text.

Yet the concentration is visible simply by tracing the sequence. The same office influenced whether an idea entered the process, how its relevance was framed, when it advanced, how discussion was summarised, who worked on the text and whether the result had achieved consensus. That does not prove misuse, and nothing in the record warrants an allegation of dishonesty by a chair or author. It does show that the design depended heavily on judgement exercised at several linked gates. Where each decision builds upon the preceding one, even conscientious discretion needs reasons and a reviewable record.

Otherwise a later declaration can conceal the cumulative effect of earlier screening, deferral or framing.

Draft 3 tried to add intellectual discipline to consensus. It stated that most Working Group decisions were made by rough consensus and expressly equated policy-development “consensus” with rough consensus. It distinguished minor from major objections. A minor objection could be discussed and, if it remained unresolved, its proponents could be asked to set it aside when the proposal’s overall advantages outweighed it. A major objection alleged serious problems for part of the community.

Participants were meant to work through such objections; rough consensus could be found when objections had been resolved or given due consideration and the benefits were judged to outweigh the disadvantages.

That formulation deserves credit because it does more than count raised hands. It recognises that an objection has content and severity. It invites the group to ask whether a concern has been answered and whether a proposal’s advantages outweigh residual costs. Properly practised, rough consensus is an engineering discipline: identify the technical invariant, listen for failure modes, revise the design and determine whether remaining objections reveal a defect or merely a preference. This can be valuable when coordinating uniqueness, accurate registry records, authorised updates, security assertions and continuity.

It is evidence about whether a technical arrangement will work.

But Draft 3 also contained a different-sounding sentence. In the same consensus section, it described consensus as everyone consenting to a decision acceptable to all participants. That sits uneasily beside rough consensus, which by definition can proceed despite unresolved objections after they have received due consideration. The available record does not establish whether the sentence was a drafting inconsistency or an intended additional test. It is therefore important not to manufacture a tidy resolution.

The text carried both formulas: a rough-consensus mechanism that tolerated some residual objection and a unanimity-like description centred on everyone’s consent.

The tension matters because the chair had the final consensus decision. Draft 3 said a show of hands was not a vote; it was one indication considered alongside discussion in the room and on the mailing list. That is sensible as far as it goes. A technical objection does not become false because its author is outnumbered, and a popular preference does not become technically necessary through applause. Yet declaring a show of hands non-binding does not answer the authorization question. It changes the evidentiary method inside the forum. It does not establish who empowered that forum to bind people who were absent, silent or never represented.

Last Call exposed the problem most sharply. It was to last at least two weeks, providing a final opportunity to comment after a consensus declaration. A final review period can catch an overlooked defect, allow participants who missed a meeting to respond and expose a difference between the chairs’ summary and the written record. Draft 3, however, expressly treated no feedback in Last Call as consensus. Under that rule, an empty evidentiary field acquired a positive procedural meaning. Silence did not merely leave the prior declaration undisturbed; silence itself counted as agreement.

That shortcut might keep a long-running discussion from repeating forever. If all known objections have already been addressed and the final text faithfully reflects the exchange, a quiet closing period may reasonably tell administrators that no new issue has appeared. But “no new issue appeared in this channel” is not the same proposition as “the affected principals consented”. Non-response can reflect satisfaction, exhaustion, lack of notice, language, time zone, staffing pressure, monitoring cost, inability to attend or a rational decision not to invest in a process one cannot influence.

The text supplied no complete denominator against which quiet could be interpreted. The sealed record has no count of all participants, meeting attendees, list readers or affected operators, and no proof that speakers were authorized to bind their organisations.

The Board gate appeared narrower. Once co-chairs sent the comprehensive lifecycle report, the Board was described as obliged to ratify unless it found a material flaw in the process or consensus determination, or considered the proposal harmful to AFRINIC in the Board’s fiduciary role. If it did not ratify, it had to publish reasons and ask the Working Group to rectify the issue. Those requirements are significant. An obligation framed with exceptions is more constrained than an unlimited choice.

Publishing reasons makes it possible to see whether the Board is correcting a process defect, identifying corporate harm or substituting its policy preference. Remand preserves the Working Group’s capacity to repair rather than allowing an unexplained institutional veto.

The implementation clock also became more coherent. Draft 3 located implementation within six months after Board ratification unless a waiver was requested. Moving the clock from Last Call consensus to ratification aligned operational preparation with the final corporate gate described by the proposal. Registry staff should not have to implement a text while the body assigned a ratification role has yet to act. A waiver mechanism also acknowledged that implementation effort can differ across policies. These are practical improvements in sequencing, not evidence that the resulting policy would possess public authority.

Alongside ratification sat a more expansive-looking emergency-variance clause. It described the Board as able to adopt number-resource policy that it considered necessary and urgent, with the Working Group then considering the measure at the next public policy meeting. The sequencing reverses ordinary deliberation: corporate action first, later community consideration. Emergencies sometimes require temporary operational measures. A registry may need to protect unique allocation, preserve data integrity, isolate fraud or keep essential services running. But the words “necessary” and “urgent” do not enlarge the institution’s mandate.

They are conclusions asserted by a private Board unless tied to defined technical triggers, evidence, duration and review.

The appeal route supplied another kind of structure. During Discussion, a complainant first raised concerns with the chairs. During Review and Concluding, a challenge generally addressed process or an allegedly incorrect consensus determination rather than the policy’s merits, except in exceptional extenuating circumstances. An appeal required support from three Working Group individuals who had participated in the discussions. It had to be filed within two weeks of the relevant chair action or decision, and a report was due within four weeks. The committee could annul a chair decision when the policy process had not been followed.

Draft 3 called the committee’s decision final.

That was a usable remedy in a limited sense. It identified who could bring a challenge, supplied clocks, specified a process-focused ground and gave the reviewing body power to undo an improper chair decision. An appeal that can only comment but cannot correct is weak; annulment provided a concrete consequence. A four-week report deadline guarded against indefinite limbo. At the same time, three participating supporters and a two-week filing window narrowed access. A person affected by an administrative result but absent from the prior discussion might lack standing under the proposed route.

A participant aware of the issue but unable to find two additional qualifying supporters before the deadline might never reach review.

Composition sharpened the institutional question. The three seats were a director selected by the Board, a Council of Elders member selected by that body and the immediate past co-chair, subject to fallback arrangements. The Council of Elders nominee chaired the committee. The preferred decision rule was consensus, with majority voting if consensus failed. This mix offered institutional knowledge: a director could understand corporate constraints, an elder might bring continuity, and a past co-chair could read the procedural record. Experience can make review faster and better informed.

Experience is not the same as independence. The committee was assembled from offices closely connected to the structure whose actions it would examine. One member was selected by the Board that received the policy; another came from an internal continuity body; the third had recently occupied the chair role. This does not prove bias in any individual. It means the design lacked a strong separation firewall. Calling the result “final” could settle an internal workflow, but it could not create judicial finality.

A private appeal committee cannot make itself a court through drafting, displace ordinary law or transform a process ruling into sovereign adjudication.

The attached guideline extended the chairs’ operational reach. It described two volunteer chairs, both engaged in consensus gathering, with one primary contact assigned for each proposal. The terms were staggered and lasted two years. Eligibility depended partly on policy-list activity, attendance at at least two public policy meetings and presence at the meeting. Chair selection for candidates present in person used a show of hands. The guideline also contained an unresolved institutional phrasing: it said the “AFRINIC community” elected chairs while also saying AFRINIC’s Nominating Committee appointed them through the stated process.

Again, the distinction is not between participation and nothing. List activity can demonstrate familiarity with the work. Previous attendance can show sustained engagement. Physical presence makes it possible to answer questions, and a show of hands offers an immediate gauge among those in the room. Two staggered chairs can share workload, reduce dependence on one person and retain continuity when one term ends. These are plausible administrative reasons for the arrangement.

They still do not identify a legal principal. People in a room can choose whom they trust to manage their discussion. They cannot, merely by being present, authorize that chair on behalf of every operator, resource holder, member, state or user in Africa. A service region is not a polity. A mailing list is not a legislature. “Community” can describe an open circle of contributors, but the label does not establish a denominator, delegation or chain of accountability. The difference between election and appointment in the guideline further underlines the need to name which institution performs which act and under whose private rules.

The guideline gave chairs moderation and access controls as well. They could reject or defer input characterised as old, minor, premature or outside scope. They could refuse the floor or restrict meeting participation in specified circumstances. Following warnings, they could suspend a disruptive user’s posting privileges for no more than thirty days, with an appeal available for suspension. Any functioning forum needs protection against repetition, disorder and abuse. Without moderation, the loudest or most persistent speaker can consume the process and make meaningful participation impossible.

Yet moderation power sits upstream of consensus evidence. A chair who classifies an objection as premature, minor or out of scope also influences the record later used to decide whether objections were adequately considered. A temporary posting suspension can be justified as forum maintenance, but it changes who can contribute during a time-limited phase. A refusal of the floor can preserve order while also removing a voice from the meeting evidence. This is why useful administrative powers require clear criteria, recorded reasons and prompt, separate review.

The point is structural, not personal: the same design concern applies even when every office-holder acts conscientiously.

The Dakar record provides the correct ending for reconstruction. AFRINIC-28 took place on 9 May 2018. Its official minutes identified Adewole Ajao and Sami Salih as session co-chairs and recorded the outcome for PDP Bis Draft 3 as “More discussion needed”. The record does not supply a complete denominator, vote total or authorization chain for the speakers. It does not show Board ratification, implementation, emergency adoption or a Draft 3 appeal. Later versions lie outside this inquiry.

The moment closes with non-advancement: a detailed proposed process had made its power map visible enough to debate, but it had not crossed into operative policy.