Summary

  • Draft 6 corrected an internal appeal path that had named the Appeal Committee as reviewer but the CEO as the terminal decision-maker. Replacing the CEO with the committee was a small textual repair with large procedural meaning.
  • The proposal distributed chairing, moderation, suspension, recall, temporary appointment, meeting-call and consensus functions across several actors. It was therefore a dependency map for keeping a private coordination forum operational, not mere administrative housekeeping.
  • Its strongest features made discretion more legible: graduated warnings, limited initial suspension, continued receive-only access, an appeal route, recall thresholds, constrained temporary chairs and recorded transitions.
  • Several fallbacks depended on AFRINIC’s Board at a moment when AFRINIC’s own records already identified a possible quorum problem. That makes resilience a legitimate question, but it does not prove that Draft 6 caused, cured or encountered later Board incapacity.
  • AFRINIC remained a private member-based technical registry, bookkeeper, service provider and coordinator. Internal consensus and “final and binding” decisions could organise its own process; they could not create sovereignty, public jurisdiction or a mandate over absent principals.
  • A safer design would make every consequential path replayable: actor, authority source, trigger, evidence, scope, duration, reason, review, fallback, null state and terminal record, with explicit degradation when an organ cannot act.

A correction that changed the decision-maker

The sixth draft arrived after an unusually precise condition had been attached to the fifth. Draft 5 was submitted on 24 May 2022. It was discussed at AFRINIC-35 on 1 June, where AFRINIC’s record says it reached rough consensus subject to the correction of a typo. Twenty days later, Draft 5 was archived as superseded and AFPUB-2020-GEN-002-DRAFT06 was submitted by Noah Maina and Alain P. Aina.

Calling the defect a typo was understandable at the level of editing. The contested passage concerned a posting suspension. Its appeal text directed the affected person to the Appeal Committee and described that committee as reviewing the co-chairs’ decision. Yet its last sentence said the CEO’s decision would be final and binding. The route began with one reviewer and terminated with another. Draft 6 replaced “CEO” with “Appeal Committee” in that terminal sentence.

The change did not merely improve grammar. An appeal is a chain of authority. It needs a reviewable first decision, a defined appellant, a destination, a standard or scope of review, and an actor whose response completes the internal process. If the chain names the Appeal Committee at its entrance and the CEO at its exit, participants cannot tell whether the CEO is meant to conduct a second review, endorse the committee, substitute a decision, hear a further appeal or simply appears by mistake. Each reading produces a different allocation of power.

Draft 6 removed that particular fork. The committee that reviewed the suspension was now also the actor whose internal decision the text described as final and binding. That made the stated route coherent, although it did not answer every question about the route. The public revision wording referred to section 3.3.10 and an adjacent sub-number while the visible corrected sentence appeared under another adjacent label. The identity of the replacement actor is clear; the subsection labelling is not fully harmonised in the sealed record.

Precision in one noun therefore solved the central contradiction without making the surrounding document immune to interpretive friction.

That is why the correction offers a useful entrance to the whole instrument. Procedures are often treated as the wrapping around substantive policy. Draft 6 showed the opposite. In a forum that relies on volunteers, mailing-list conduct, public meetings, rough consensus and corporate fallbacks, procedure is the execution layer. It decides whether a warning is merely advice, whether a silence counts as consent, whether a temporary chair may determine consensus, whether a recall is admissible, and whether a missing office-holder causes substitution or stoppage. A stray decision-maker is not a cosmetic blemish in such a system.

It is a branching error in the control path.

Six drafts were evidence of work, not guilt

AFRINIC identified Draft 6 as the sixth version of a proposal family that began with a posting to the RPD list on 22 July 2020. Draft 4 was discussed at AFRINIC-34 on 17 November 2021. The official record later said that version did not achieve rough consensus and returned to the mailing list. By the time Draft 5 reached the next meeting, the text had undergone repeated exposure to discussion and revision.

It is tempting to make the number six do more analytical work than it can bear. Six versions do not, by themselves, prove obstruction, capture, incompetence or bad faith. A difficult procedure can legitimately require several rounds because it must behave reasonably under ordinary disagreement and exceptional vacancy. Revision can be a sign that authors listened, that participants found edge cases, or that the forum was trying to replace unwritten custom with testable instructions. Equally, repeated amendment can reveal that apparently modest administrative language contains many linked choices.

The defensible conclusion is narrower: the details mattered enough to keep revising.

The proposal itself said the Policy Development Process lacked specific working-group procedures and operational rules, and that the absence had led to interpretation problems in administration and moderation. Its stated remedy was to amend section 3.3 of the Consolidated Policy Manual and make section 3.5.3 obsolete. The intended object was not an Internet-number allocation rule. It was the behaviour of the Policy Development Working Group: how it would be administered, how its co-chairs would be chosen and held to account, and how participation would continue during conflict or vacancy.

The chronology after 21 June also requires restraint. Draft 6’s page records Last Call from 23 June to 14 July. It then records a consensus announcement, a ratification report sent to the Board and withdrawal by the authors on 28 July, all on the same date. The draft was archived on 1 August. Those are process states, not a completed theory of what occurred. The sequence does not establish that the Board ratified or considered the proposal, that any clause entered operation, that the authors withdrew for a particular reason, or that the same language later appeared in another instrument.

The coincidence of consensus announcement and withdrawal naturally raises a question, but it does not answer it. Withdrawal might have followed any number of considerations, and none should be supplied without evidence. Nor does the phrase “ratification report sent” prove the act of ratification. A transmission is not a decision. An archived proposal is not necessarily an implemented rule. Draft 6 is analytically important because of what it specified and what its milestones record, not because an uncertain endpoint can be converted into a convenient success or failure narrative.

The forum as a dependency map

Draft 6 described the working group as open to interested individuals and administered primarily by two volunteer co-chairs. Those two propositions created the central design tension. Openness could draw technical knowledge, operational experience, dissent and scrutiny from a wide field. Administration still required identifiable people to prepare meetings, moderate discussion, summarise views, determine proposal status, supervise Last Call and pass a ratification report towards the Board.

The co-chairs were therefore neither ornamental hosts nor unbounded rulers. They sat at a junction between participation and record. Before a meeting, they had organisational duties. During it, they had to conduct discussion. Afterwards, they had to preserve and interpret the decision trail. Their ability to say whether rough consensus existed gave them influence over a proposal’s progress, but that influence arose from a private procedural assignment. It was not legislative power, and the meeting did not become a legislature because a policy text was debated there.

Other actors entered for particular contingencies. The Appeal Committee reviewed a posting suspension. A recall committee could draw one person through the Council of Elders channel, one through the Governance Committee and another through community channels. The Board called Public Policy Meetings and appeared as a fallback appointment actor when the working group could not fill specified chairing vacancies. The CEO appeared in the predecessor draft’s inconsistent terminal appeal sentence, then ceased to occupy that stated role after the correction.

None of this produces a single ladder on which every actor is simply above or below every other actor. It is a network of assigned tasks.

That distinction matters. A Board fallback in one appointment path does not imply that the Board owns every co-chair judgment. An Appeal Committee’s terminal role in a suspension appeal does not turn it into a general court. The Governance Committee’s possible contribution to a recall body does not make its advice legislation. The Council of Elders channel supplies a route to committee composition, not mystical senior authority. A procedural map must be read by function, trigger and scope, not converted into a vague hierarchy of prestigious names.

The map also contained hard stopping points. If both co-chairs were absent at a Public Policy Meeting, the working group would first try to appoint a temporary chair. If it could not do so, the Board could appoint one. If the Board failed to appoint, the meeting would be adjourned. That last instruction is as important as the substitutions before it. It recognises that not every absence should be cured by whoever happens to be available. Sometimes a process should stop rather than allow authority to migrate informally.

The strongest case for writing the rules down

The most persuasive reading of Draft 6 starts with the practical vulnerability of a forum governed mainly by custom. Custom works smoothly when participants share expectations, chairs are present, tempers are moderate and institutional memory is intact. It becomes unreliable when the very question in dispute is what the custom permits. Then the people with the strongest recollection, longest tenure, loudest voice or greatest access can appear to own the rule.

Written procedure can reduce that danger. It tells a participant what conduct may prompt intervention. It requires a sequence before a severe step. It can separate the person who makes an initial decision from the body that reviews it. It tells would-be recall organisers how many supporters they need and what connection those supporters must have to the forum. It prevents a temporary chair, appointed only to keep a meeting moving, from taking the more consequential step of determining consensus. It can preserve a suspended participant’s ability to receive messages, maintaining visibility into the process even when posting is restricted.

These are not trivial gains. A volunteer-run technical forum cannot assume an inexhaustible supply of trusted office-holders. Nor can it leave every dispute to personal negotiation. If the work affects the reliability and predictability of a registry’s coordination services, participants and operators benefit from knowing how the forum will behave before a crisis begins. LARUS’s operator-continuity perspective is relevant in this limited sense: uncertainty in registry governance can impose diligence and continuity costs on people who depend on predictable administration.

It does not show that Draft 6 caused an outage, changed a route, deprived a customer of a resource or generated a financial loss.

The benign case also explains why narrow limitations inside Draft 6 deserve attention. An initial posting suspension was capped at no more than 30 days. The individual could continue to receive mailing-list messages. A temporary meeting chair could guide discussion and provide minutes but could not decide consensus on policy proposals. Recall required more than an aggrieved individual. These constraints indicate an attempt to keep necessary administration from silently accumulating broader powers.

The correct response to that benign case is not to reject procedure as bureaucratic theatre. It is to demand more of procedure because it is useful. Every action with consequences should be reconstructable later. A warning should have a trigger and evidence. A suspension should have scope, duration and reason. An appeal should name its reviewer and the limits of review. A fallback should say what happens if the fallback actor is missing. A decision described as final inside the process should still produce a record and remain bounded by contract, corporate constitution and ordinary law.

Draft 6 was strongest where it approached that standard. It was weakest where an institutional dependency was named without a fully specified degradation path, or where the language of finality risked sounding larger than the private process it served.

Appointment by consensus, ballot as a last resort

The proposal envisaged two co-chairs serving staggered two-year terms. The working group would appoint them by consensus, using a secret ballot only as a last resort if consensus could not be reached. Staggering aimed at continuity: one office-holder’s term could turn over without simultaneously erasing all chairing experience. Consensus sought a broadly acceptable administrator rather than a winner sustained only by a narrow count. The ballot offered a terminal device when discussion could not settle the selection.

Each element addressed a different failure mode. Simultaneous terms risked a complete loss of memory. An immediate ballot could transform an administrative office into a factional prize before the group had tested whether objections could be answered. Consensus without any endpoint could allow a small number of persistent objectors to keep an office vacant. The combined design tried to hold continuity, deliberation and closure together.

Yet “consensus” is not self-executing. Someone must frame the question, determine when discussion has been sufficient, distinguish a reasoned objection from repetition, and record why the conclusion reflects rough agreement rather than exhaustion. When the office being selected is the office that ordinarily interprets consensus, the forum faces a circularity. The selection process needs a conductor before it has selected the conductors.

Draft 6’s wider appointment and temporary-chair provisions can be read as attempts to deal with that circularity, but the dependency remains worth making explicit. Who ascertains that consensus cannot be reached? What record shows the transition to secret ballot? Who controls eligibility, the voter roll, the count, challenges and publication of the result? The sealed facts establish the consensus-first, ballot-last structure, not every operational answer to those questions.

Heng Lu’s participation-not-mandate principle supplies an essential limit. Open participation can produce evidence, practical expertise, objections and warnings. It can make an internal administrator better informed and more accountable to the people who use the forum. Attendance does not manufacture a public mandate, and consensus among those present cannot bind absent principals merely because the meeting called itself multi-stakeholder. The co-chairs would administer AFRINIC’s private policy process. They would not become representatives of Africa in a constitutional sense.

This limit does not make the appointment method pointless. A private association may sensibly invite a broad technical constituency to help choose the people who moderate its forum. The result can be legitimate within the terms of that forum without becoming sovereignty. Precision about the boundary strengthens the process: participants know they are choosing custodians of a limited coordination function, not conferring a general power that no meeting has authority to give.

Moderation needed an evidence trail

Moderation is where procedural abstractions meet a person’s ability to speak. Draft 6 allowed posting suspension after graduated warning. The initial period could not exceed 30 days, the affected person could still receive list messages, and an appeal was available. These choices reveal an effort to distinguish protection of the discussion channel from exclusion of the individual from all knowledge of it.

Graduation matters because a warning gives the participant an opportunity to change behaviour and gives the moderator an opportunity to explain the rule. A capped initial suspension limits the first deprivation. Receive-only access reduces the risk that a person disappears into an informational blind spot while others discuss the suspension or the underlying policy. Appeal creates institutional distance between the co-chairs’ first decision and the review of that decision.

But a sequence is fair only if its inputs can be checked. “Warning” should mean more than a private recollection that the chair once expressed displeasure. The record should identify the communication, the conduct at issue, the provision applied, the requested correction and the consequences of repetition. If a later suspension relies on repeated behaviour, the decision should connect the new incident to the previous notice. Otherwise graduation becomes a label placed on accumulated discretion rather than a constraint on it.

Duration requires the same care. An initial cap of 30 days is meaningful, but a repeated-behaviour clause can become an indirect route to open-ended exclusion if extensions lack a fresh trigger, separate reasons and a new terminal date. A safe process would never let “repeat” serve as a perpetual status. Each extension would need current evidence, a defined period and another opportunity for review. The public record need not publish sensitive personal detail indiscriminately, but the affected person and reviewer need enough evidence to reconstruct the chain.

Scope matters too. A posting suspension should restrict the particular ability necessary to protect the channel, not become a general sanction on registry services, membership rights, number-resource records or participation in unrelated settings. Draft 6’s receive-only provision pointed towards narrowness. The private-authority boundary requires that narrowness to be explicit. A co-chair moderates a private discussion space under assigned procedure. A co-chair is not a police officer, prosecutor or punisher, and cannot turn a mailing-list decision into a public disability.

The appeal correction then becomes the hinge of the safeguards. If the Appeal Committee reviews the suspension, its record should state what it considered: compliance with procedure, reasonableness of scope, sufficiency of evidence, proportionality of duration, or some defined combination. The phrase “final and binding” can sensibly mark the end of the internal ladder. It cannot transform the committee into a public-law court or prevent a competent court from exercising jurisdiction under ordinary law. Internal closure is a routing instruction, not a source of sovereignty.

That distinction also explains why the Draft 5 wording was dangerous even if everybody suspected a typo. If the CEO appeared at the end, a participant could reasonably ask whether a corporate executive could replace or override the committee’s review. The ambiguity would be sharpest precisely when trust was lowest. Draft 6’s replacement aligned the terminal sentence with the named appellate body. It made the internal route more intelligible, while leaving the route’s authority as private as before.

Recall as a protection against both impulse and entrenchment

The proposal did not make recall available on a lone request. It required the requester and at least ten supporters from ten organisations. It also required at least one year of mailing-list subscription and attendance at one Public Policy Meeting in the preceding two years. These thresholds attempted to establish that the challenge came from people with continuing exposure to the forum rather than a momentary influx organised around a single grievance.

There is a benign logic to those conditions. Recall can protect a working group from an office-holder who has lost confidence or cannot perform. Yet an effortless recall mechanism can make chairs vulnerable to constant tactical campaigns, encouraging them to placate the most mobilised bloc. Requiring breadth across organisations and some record of participation imposes friction. It asks challengers to show that concern extends beyond one institutional interest and that supporters have observed the process over time.

Friction can also exclude. A subscription-duration rule favours established participants over people affected by recent decisions. A meeting-attendance condition may privilege those able to travel, obtain visas, take time away from work or secure institutional support. An organisation-count rule can look broad while multiple supporters still share an aligned interest. The thresholds are therefore evidence of intended screening, not proof that the resulting group represents the wider region.

The recall committee was to contain three people drawn through different channels: the Council of Elders, the Governance Committee and community channels. It had four weeks to decide, and the result was described as final and binding within the process. Mixed composition could reduce the danger that one office or faction controls the inquiry. A deadline could prevent recall from leaving the co-chair and the working group in indefinite limbo.

Still, labels for channels do not guarantee independence. A robust recall path would disclose how each person was selected, conflicts of interest, the allegations accepted for review, evidence supplied, the co-chair’s opportunity to respond, the decision standard and the reasons for the outcome. “Final” should end the internal process, not end accountability for whether the process followed its own rules. The committee’s limited function would be to decide a private office-holding question. It would not adjudicate public rights.

The four-week deadline also needs a null state. What if the committee cannot be constituted? What if a member withdraws? What if the deadline expires without a decision? Silence should not automatically remove the co-chair, clear the co-chair, or let an interim arrangement continue indefinitely. Each outcome would allocate power. A mature control plane would say which status remains in force, for how long, and what record marks the failure.

Vacancies exposed the Board dependency

Recall does not end when an incumbent leaves. Someone must keep the working group functioning. Draft 6 said that if the working group could not appoint interim co-chairs after recall, the Board was to appoint them within two weeks after the lack of consensus was ascertained. The Board thus served as a fallback for a forum that had failed to fill its own temporary offices.

The logic is understandable. A corporate Board is a recognisable organ, and using it can be preferable to allowing an ad hoc participant to seize the chair. A two-week deadline limits vacancy. The fallback is also bounded in subject: it concerns interim co-chair appointment, not a transfer of every working-group function to directors.

But the clause assumes that the Board exists in an operational form capable of acting. That assumption was already under visible strain before Draft 6 was submitted. AFRINIC’s signed Board minutes record that, on 2 March 2022, one director warned of a possible post-AGMM quorum problem if a majority of directors stood for election at the same time. Another director disputed the forecast. The record therefore establishes both the warning and the disagreement; it does not allow the warning to be presented as an uncontested prediction.

The same minutes recorded Resolution 202202.677, which addressed an election-sequence defect involving two seats that had been filled casually and asked the Governance Committee for workable recommendations by the end of April. On 28 April, the Governance Committee publicly identified the inability to raise Board quorum when many seats were open as a governance risk and recommended a three-seat sequence. Its communiqué was advisory. It was not an enacted bylaw, a court judgment or proof of how subsequent events had to unfold.

These records make the Board fallback analytically material because they show that institutional participants were already discussing the availability of the very organ on which the fallback depended. They do not show that Draft 6 caused the risk, that the proposal should have cured it, that a fallback was ever invoked, that the Board failed to perform it, or that the authors withdrew because of it. The later BTW account of a Board without quorum establishes why the dependency would acquire serious institutional stakes. That later story owns the collapse narrative.

Draft 6’s story stops at the earlier design question: how should a procedure behave when a named corporate organ may be unable to act?

A dependency-tested clause would not merely name the Board. It would state the minimum decision capacity required, the evidence that the working group had failed to appoint, who sends the referral, when the two-week clock begins, what constitutes a valid Board appointment, and what occurs if no valid appointment is made. It would also forbid the fallback from expanding its remit. A Board filling an interim chair vacancy should not thereby acquire the co-chair’s consensus function or the working group’s policy voice.

The most important missing concept is safe degradation. When an organ is unavailable, three broad responses are possible: transfer its function, narrow the activity that can continue, or stop. Transfer is often treated as the practical answer, but it carries the greatest capture risk. A substitute chosen for convenience can acquire a role it was never designed to hold. Narrow continuation may be safer: administrative notices can proceed while substantive consensus determinations pause. Sometimes stopping is safest of all.

The empty-chair rule understood restraint

Draft 6’s treatment of a Public Policy Meeting with both co-chairs absent demonstrates that stoppage was not alien to its design. The Board called the meeting. If both co-chairs were unavailable, the working group would try to appoint a temporary chair. If it failed, the Board would try. If the Board failed too, the meeting would be adjourned.

The temporary chair’s powers were explicitly limited. The person could lead discussion and provide minutes but could not determine consensus on policy proposals. That is a significant separation. It allows participants to exchange views and preserve a record while reserving the decision that moves a proposal through the process. Presence at the lectern does not automatically grant the full authority of the missing office.

The design contains a sound principle: emergency continuity should conserve the forum, not silently rewrite its constitution. A temporary chair can perform reversible, record-producing acts. A consensus determination is harder to reverse and can change the state of a policy proposal. By withholding that function, Draft 6 kept the substitute thinner than the ordinary role.

The adjournment condition extends the principle. If neither the working group nor the Board can produce a temporary chair, the meeting stops. That may be inconvenient, expensive and frustrating. It is nevertheless preferable to an unrecorded transfer of control to the loudest participant, staff member or available dignitary. Institutional resilience is not the ability to keep every activity going at any cost. It is the ability to fail without manufacturing authority.

Even this relatively careful path raises dependency questions. If the Board itself called the meeting but later lacked capacity to appoint a temporary chair, who records the failure? Can discussion continue informally without being treated as a Public Policy Meeting? When is the adjourned meeting reconvened, and by whom? Does the same agenda carry over? Can the written record distinguish useful consultation from a formal stage in the Policy Development Process? The facts provided do not answer all of these questions. They illustrate why null states require design rather than improvisation.

The temporary-chair clause also clarifies what consensus should not mean. It is not simply the mood in a room that any facilitator can announce. Within AFRINIC’s private procedure it was a specific status judgment assigned to authorised co-chairs and linked to further stages such as Last Call and a Board report. That assignment can be criticised, reviewed and constrained, but it cannot safely be assumed by anyone who keeps the microphone working.

Consensus records need receipts

Draft 6 assigned the co-chairs tasks across the life of a proposal: guiding discussion, summarising positions, deciding status, monitoring Last Call and submitting a ratification report to the Board. That sequence converts conversation into institutional state. A proposal moves from discussion to an announced conclusion, from a conclusion to Last Call, and from Last Call towards a report.

Each movement should leave a receipt. A useful consensus record would identify the exact version considered, the question presented, material objections, the co-chairs’ answer to each objection, changes made, unresolved concerns, the date and channel of announcement, the Last Call interval, developments during that interval and the basis for the final status. Without such a ledger, “rough consensus” can become an assertion whose supporting discussion is scattered across recordings, list archives and recollections.

Receipts do not turn consensus into a vote or require unanimity. They make judgment reviewable on its own stated terms. A reasoned objection can fail to prevent consensus, but readers should be able to see that it was heard and why it did not control. Repetition can be distinguished from a new argument. A correction can be tied to the exact version that incorporates it. The Draft 5-to-Draft 6 episode makes the importance of versioning unusually vivid: the meeting-stage result was conditional on correction, and the successor version changed the identity of the terminal appeal actor.

The public milestone list provides a skeleton of such a ledger. It records dates for submission, Last Call, consensus announcement, report transmission, withdrawal and archive. Yet a skeleton is not the full decision trail. Why did consensus announcement and withdrawal occur on the same day? Did the Board receive the packet in an administratively complete form? Was it considered? Did withdrawal end consideration automatically? Was any clause carried elsewhere? Those questions remain open.

The right response to missing answers is not suspicion presented as fact. It is to mark the uncertainty and define the evidence that would resolve it. A versioned public decision ledger should connect objections, responses, amendments, announcements and terminal state. It should distinguish “sent” from “received”, “received” from “considered”, and “considered” from “approved”. It should not use an archive label to imply implementation or failure unless the governing rule gives that label such an effect.

This is where repeated drafting most clearly resembles infrastructure. A system that cannot identify which version a decision concerns cannot reliably execute its own rules. The distinction between Draft 5 and Draft 6 was one actor name, but an appeal brought under the wrong version could reach a different terminus. Version control is not administrative fuss when it determines who can decide.

A bookkeeper needs procedure, not a throne

AFRINIC’s work can be operationally important without being sovereign. NRS’s account supplies the relevant institutional frame: AFRINIC is a member-based regional Internet registry whose services support the administration of number resources. It keeps registry records, coordinates private processes and provides technical services within a wider Internet system. Those functions can matter greatly to operators, networks and users.

Importance does not convert the registry into a state. AFRINIC is not a legislature because it publishes policy proposals. It is not a regulator because participants call a decision binding inside its process. It is not a police force or prosecutor because co-chairs may moderate a mailing list. It is not a confiscator because registry entries concern scarce identifiers. It is not a public-law adjudicator because an Appeal Committee reviews an internal suspension.

Heng Lu’s bookkeeper principle is useful precisely because it refuses a false choice between insignificance and sovereignty. A bookkeeper may perform an indispensable service. Uniqueness of records, consistency of registration and dependable coordination can be essential to operational planning. The service becomes dangerous when custody of the ledger is rhetorically transformed into ownership of the underlying world, or when a private process treats its convenience as jurisdiction over everyone affected by Internet infrastructure.

Draft 6 should be understood on the service side of that boundary. Its co-chair rules could make the forum less arbitrary. Its moderation process could protect the usability of discussion. Its appeals and recalls could constrain office-holders. Its vacancy rules could preserve continuity. All of these are legitimate objects of private institutional design. None authorises AFRINIC to bind a person who never agreed to its process in the way a sovereign law binds within public jurisdiction.

The phrase “final and binding” therefore needs a deliberately modest reading. It can say: there is no further appeal inside this particular procedural ladder. It cannot say: no competent court may examine contractual, corporate or other legal rights. It cannot create the procedural protections or democratic authority of public adjudication merely by borrowing the vocabulary of judgments. The same applies to Board actions described as final in appointment paths. Finality closes an internal route; it does not create immunity.

The distinction improves drafting. If a rule-maker knows that its authority is limited, it can specify the exact administrative question being decided. It can avoid grand language and publish reasons. It can preserve an external legal boundary without trying to draft around it. Thin authority is easier to audit because the institution is not pretending that every disagreement belongs inside its own system.

Participation supplies knowledge, not a regional mandate

The working group’s openness was a practical asset. Network engineers, resource holders, researchers, civil-society participants and other interested individuals can detect consequences that a small corporate office might miss. They can bring evidence from different operating environments, challenge ambiguous language and warn about failure modes. A public list creates a larger surface for criticism than a closed committee.

But an open door is not a constitutional electorate. The people who participate are self-selecting and differently resourced. Some organisations can fund repeated attendance and sustained list activity; others cannot. Language, bandwidth, time zones, travel, professional incentives and confidence in institutional culture all affect who appears. A requirement tied to list tenure and meeting attendance may ensure familiarity while also deepening those selection effects.

For that reason, rough consensus should be described as a decision method for AFRINIC’s private policy forum, not the sovereign will of an African Internet community. Participants can authorise nothing beyond the authority they possess. They can advise, object, offer expertise and help the registry set terms within its lawful private role. They cannot confer public powers that neither they nor AFRINIC hold.

This is not an argument for closing the forum. It is an argument for honest institutional language. Claims of representation should identify who participated, under what conditions and with what legal relationship to the decision. Consensus records should preserve minority objections, not erase them under a collective noun. The broader the rhetoric of “the community”, the more important it becomes to show the actual composition and limits of the group that spoke.

Draft 6’s procedural precision could support that honesty. Defined appointment rules tell readers how administrators were selected. Recorded moderation shows how the discussion surface was managed. Versioned consensus receipts show which objections entered the decision. Recall rules indicate how office-holders could lose confidence. These mechanisms do not create a mandate, but they can make a private process more accountable and epistemically useful.

Operator continuity without a fictional outage

Registry governance can feel remote from day-to-day network operation until an institutional dependency fails. Operators plan around stable records, predictable service windows, clear points of contact and intelligible change processes. When internal authority becomes ambiguous, they may spend more time on diligence, contingency planning, documentation and escalation. That burden can be real even before a technical incident occurs.

LARUS’s continuity analysis helps explain why this class of governance deserves attention. It does not establish a factual chain from Draft 6 to routing instability, customer harm, resource loss or financial damage. No such claim follows from the sealed evidence. Nor does the proposal’s withdrawal prove that an operator experienced a failure. The relevance lies in exposure: procedural ambiguity can add uncertainty around a service on which operators depend.

The Draft 5 appeal defect is a small example of that exposure. If two actors can plausibly be read as terminal reviewer, a dispute may take longer, consume more staff attention and undermine confidence in the moderation channel. The Board fallback is a larger example. If a vacancy rule depends on an organ that cannot validly act, the working group may be unable to fill the office that records consensus. The consequences still need evidence; the dependency itself is visible in the text.

Good procedural infrastructure reduces the number of facts an operator must guess. It says which version is current, who can issue a decision, how long it lasts, where review goes and what happens during institutional absence. It separates discussion continuity from decision authority. It provides a terminal record even when the outcome is “no valid actor was available”. Such clarity supports planning without promising that governance can eliminate every operational risk.

The later Board-without-quorum context is valuable because it shows why a corporate dependency can become more than a drafting curiosity. Yet importing the whole later crisis into June 2022 would distort the commission. The contemporaneous evidence is sufficient: a possible quorum problem was raised, disputed and then identified as a governance risk by an advisory committee. That is enough to test Draft 6’s fallback design. It is not enough to write a causal history of collapse.

The safer counterfactual: a replayable control plane

Imagine Draft 6 not as a set of paragraphs but as a control plane that must execute during disagreement. Every consequential route would carry eleven fields: actor, authority source, trigger, evidence, scope, duration, reason, review, fallback, null state and terminal record. The point would not be to mechanise judgment out of the forum. It would be to ensure that judgment cannot hide its dependencies.

For a moderation warning, the actor would be the authorised co-chairs or the specifically designated chair. The authority source would be the numbered working-group rule. The trigger would be defined conduct, not general annoyance. Evidence would include the relevant posts and prior notices. Scope would be limited to the discussion channel. Duration would describe how long the warning remains relevant. The reason would connect the evidence to the rule. Review might begin only if a later suspension relied on the warning. The fallback would address a conflicted or absent co-chair.

The null state would preserve posting rights if no authorised actor could decide. The terminal record would show notice and receipt.

For a suspension, the same fields would become stricter. The trigger would require the completed warning sequence or a narrowly defined urgent exception. The evidence bundle would be frozen when the decision was issued. Scope would specify posting, not receiving, and exclude unrelated registry services. Duration would name start and end times and never rely on an indefinite “until further notice”. The written reason would explain necessity and proportionality. Review would go to the Appeal Committee, not a stray executive actor.

If the committee could not sit, the safe fallback might be expiry of the suspension at the initial cap rather than automatic extension. The terminal record would distinguish affirmed, varied, expired and revoked outcomes.

For recall, the actor receiving the request would verify the requester, ten supporters, ten-organisation spread, subscription history and recent meeting attendance. The authority source would state why those criteria control. Evidence would be split between threshold evidence and the merits of the complaint. Scope would be the office, not the individual’s other relationships with AFRINIC. The four-week duration would include rules for replacing a conflicted committee member. Review might be limited, but the result would still have reasons.

If the committee could not form, the null state would say whether the incumbent remains temporarily in office and what functions, if any, are narrowed.

For co-chair appointment, the control plane would define who facilitates the consensus process before co-chairs exist. It would record candidacies, eligibility, objections and the test for moving to secret ballot. The ballot would have a defined electorate and challenge path. If no result emerged, an interim arrangement would carry only the powers necessary to preserve the list and meeting record. No interim appointee would gain a longer mandate through delay.

For a post-recall vacancy referred to the Board, the trigger would be a public record that working-group consensus had failed. The referral would identify the exact interim office and two-week period. Board capacity would be verified under the corporate constitution. The appointment instrument would be published. If the Board lacked quorum or did not act, the null state would not allow staff, a committee or an individual director to inherit the power by implication. Instead, substantive proposal-state decisions would pause while reversible administrative functions continued under a pre-authorised custodian.

For a meeting with both co-chairs absent, Draft 6 already supplied much of the model. The working group tries to appoint a temporary chair, then the Board; if neither succeeds, adjourn. The counterfactual would add a record of each failed step, a precise boundary around informal discussion, a reconvening rule and a prohibition on treating minutes from an informal exchange as a consensus determination. The temporary chair’s existing inability to determine consensus would remain.

For consensus itself, the actor would be the properly appointed co-chairs. The trigger would be completion of the specified discussion stage. Evidence would be a versioned objection-and-response ledger. Scope would be the status of the exact proposal version. The reason would explain treatment of material objections. Review would include Last Call and any formal appeal route provided by the governing process. If co-chairs were absent or conflicted, no substitute would infer consensus unless the rules expressly granted that narrow function.

The terminal record would distinguish no consensus, rough consensus, Last Call, returned for revision, report transmitted, withdrawn, archived, considered and ratified.

This control plane would be deliberately thin. It would not empower AFRINIC to decide public rights. It would not turn the working group into a parliament. It would not claim that technical coordination overrides contracts, corporate law or courts. Its ambition would be narrower and more valuable: make the registry’s own private procedure predictable, limited and inspectable.

Safe degradation is a governance capability

Institutional documents often treat continuity as synonymous with substitution. If the first actor is absent, a second acts; if the second is absent, a third appears. Every extra fallback seems to make the system more resilient. In reality, a long chain can make the system less safe if each substitute is weakly authorised or gains powers beyond the original role.

Safe degradation asks a different question: what is the minimum activity that can continue without corrupting the process? A mailing list can remain readable even when posting moderation is under appeal. A meeting can collect views without determining consensus. Staff can preserve records without choosing a policy outcome. A proposal can wait without being deemed rejected or approved. These are reduced modes, not total paralysis.

Null states must therefore be written as carefully as active powers. “No decision” should not silently favour one side. If a suspension appeal cannot be heard, continuation of the restriction favours the moderator; automatic reversal may expose the channel to the conduct at issue. A time-limited initial decision followed by expiry unless an independent reviewer acts distributes that risk transparently. If a recall committee misses its deadline, the rule should specify whether the office-holder continues and under what temporary limits.

If the Board cannot appoint an interim co-chair, the working group should know which administrative functions survive and which policy states freeze.

Safe degradation also protects against authority laundering. Without an explicit null state, staff or influential participants may act “just this once” to preserve momentum. The emergency act then becomes precedent, and precedent becomes an unwritten office. A rule that requires adjournment or pause makes the absence visible. Visibility creates pressure to repair the legitimate institution instead of normalising the substitute.

The Board dependency in Draft 6 makes this point especially sharp. The Board was a formal corporate organ, not a magical reserve of authority. If it could act validly, a narrow appointment fallback might be useful. If it could not, the solution was not to pretend that the Chief Executive, Governance Committee, Council of Elders or an individual director automatically became the Board. Each actor had its own assigned role. Resilience required a lawful, pre-specified reduced mode.

This is also the answer to the fear that procedural exactness causes rigidity. Exact rules need not demand that all activity stop upon the slightest defect. They can distinguish reversible from irreversible action, administration from adjudication, consultation from decision and internal finality from public jurisdiction. The system becomes flexible within declared limits, rather than flexible according to whoever controls the moment.

What the surviving record still cannot tell us

Draft 6’s public history leaves several consequential questions unanswered. The first is documentary. The revision note points to neighbouring subsection labels in a way that is not fully consistent with the visible placement of the corrected appeal sentence. The actor correction is clear, but a clean consolidated instrument would be needed to remove all numbering doubt.

The second is chronological. Consensus was announced, a ratification report was recorded as sent to the Board, and the authors withdrew the proposal on 28 July 2022. The public milestone list does not explain the same-day withdrawal. It does not show whether the Board received or considered the packet, and it does not establish ratification. The 1 August archive state records archival, not implementation or legal effect.

The third concerns inheritance. The sealed record does not show whether any Draft 6 clause later entered another instrument. Similar language elsewhere would need version comparison and adoption evidence; resemblance alone would not prove incorporation. Nor can later practice be assumed to have invoked a Draft 6 fallback if the proposal never became operative.

The fourth concerns execution evidence. How were progressive warnings recorded? Did an affected participant receive the evidence on which suspension depended? Could repeated-behaviour extensions become indefinite? What standard would the Appeal Committee apply? What review, if any, existed for a Board appointment described as final and binding? The draft text gives enough structure to identify these questions, not enough evidence to answer them all.

The fifth concerns failure capacity. How would an interim co-chair be appointed if the Board lacked quorum? Who would formally record the Board’s inability to act? Which decisions would pause, and which administrative services would continue? Draft 6’s adjournment rule offers one safe-stop pattern, but the sealed record does not establish a comprehensive degraded mode.

The sixth concerns the public receipt for consensus. A complete record would connect objections, responses, the Draft 5 condition, the Draft 6 correction, Last Call, the consensus announcement, report transmission, withdrawal and archival. The milestone page preserves important states, yet it does not resolve the reasons and hand-offs between all of them.

These are watchpoints, not invitations to invent scandal. The evidentiary discipline should be symmetrical. Official AFRINIC material proves what AFRINIC published, drafted or recorded, not that its institutional language about community or consensus is normatively correct. NRS, Heng Lu, LARUS and BTW supply valuable analytical frames, not proof that they drafted, adopted, applied or adjudicated the proposal. None should be used to smuggle speculation into the chronology.