Summary

  • AFRINIC’s February 2011 implementation analysis reduced the ambitions behind its replacement policy-development process to three goals and said that only one—the dispute mechanism—had been explicitly satisfied. Better documentation still depended on operational practice, while equal weight for mailing-list and meeting comments remained in the hands of the chairs.
  • That assessment matters because AFPUB-2010-GEN-004 had already supplied a unusually candid inventory of the process it was replacing: no rule for absent co-chairs, no stated route for challenging chair action, list work that could be reopened by an uninformed meeting room, unequal practical weight between remote and in-person comments, and inadequate records of procedure, approval and implementation.
  • GEN-004 was a genuine institutional repair. It made participation and stages more visible, added appeal and recall machinery, constrained some Board discretion, and turned several implicit customs into written procedure. Yet the decisive translation of dispersed comment into “consensus” still ran through private administrators whose judgments were not made reproducible by the text itself.
  • The lesson is bounded. AFRINIC may keep an accurate, unique and operationally reliable registry, and its private process may organise work within that function. Openness does not make participation representation, however, and an internal consensus label cannot confer legislative, regulatory, adjudicative or sovereign authority over operators or absent principals.

The inconvenient sentence came after adoption

The most revealing line in the history of AFPUB-2010-GEN-004 did not appear in the document’s sales pitch. It came after the new process had been approved. On 2 February 2011, an AFRINIC staff implementation analysis looked back at the replacement policy and separated its intended repair into three parts: provide a dispute mechanism; document procedure, approval and implementation; and give online comments the same weight as comments made at the Public Policy Meeting. Its conclusion was disarmingly narrow. Only the dispute mechanism, it said, had been explicitly satisfied by the policy.

The remaining two objectives were not dismissed. The note proposed a substantial communications programme to produce the missing record: online histories of versions and events, change logs with reasons, analyses and staff comments, reports by the chairs, meeting minutes and published Board decisions. But those measures were administrative practices surrounding the policy, not direct requirements that the policy had fully settled. On parity between the mailing list and the room, the note was plainer still. That objective would be left to the way chairs conducted face-to-face deliberations.

This was an internal implementation reading, not an accusation from outside and not proof that online comments were in fact ignored. Its value is more exact. It establishes what the authoring institution’s own staff record said the text did and did not directly solve. The distinction cuts through a familiar ambiguity in private technical governance: making a process open to more people is not the same as specifying how their evidence affects the outcome. Access can be formalised while influence remains discretionary.

That is why the 2011 sentence sends the reader back to 2010. GEN-004 was a replacement draft for AFPUB-2008-GEN-001, the policy then identified as AFRINIC’s existing policy-development process. The replacement did not merely promise improvement. Its drafting record enumerated failures in the incumbent system. The rulebook was auditing its predecessor. The implementation note then audited the repair. Read together, those two acts reveal both a serious effort at institutional learning and the point at which that learning stopped.

For an operator, the distinction is not academic. A private registry’s procedure can shape the internal rules it applies. If remote objections are compressed on the way into a meeting, if no durable record connects a proposal version to the reasons for its approval, or if a dispute depends on the same practical control surface as the original decision, later costs can fall on those who were neither present nor represented. The sealed record supplies no figures for such costs, no operator denominator and no transaction-level proof. It does supply the mechanism by which procedural compression can become an operational rule.

A defect inventory, not a retrospective theory

On 23 March 2010, S. Moonesamy posted version 2 of the proposal to the RPD mailing list. The message expanded the reason for change into a written diagnosis. Its importance lies in its specificity. It did not offer a general meditation on rough consensus. It identified concrete omissions and mismatches in the process in force.

First, the existing PDP did not say what should happen if the co-chairs could not attend a policy meeting. This was a continuity defect. A process that assigns critical steps to named officeholders needs a bounded substitute rule; otherwise the absence of an individual can delay work or invite improvisation by another institutional actor.

Second, it did not specify what a person could do when disagreeing with an action taken by a co-chair or the co-chairs. This was a dispute-path defect. It did not establish that a chair had abused power, nor did it convert a procedural disagreement into a legal controversy. It showed that the process lacked an articulated internal route for contesting the conduct of its own administrators.

Third, issues discussed and apparently resolved on the mailing list could be opened again at the meeting because people in the room might not know what had already happened online. This was a memory and transmission defect. The process had two participation venues but no sufficiently reliable bridge between them. A public archive by itself could preserve comments, yet preservation did not ensure that the comments travelled into the moment when consensus was assessed.

Fourth, procedure, approval and implementation were insufficiently documented. This was not one missing minute or one broken page. It was a problem of reconstructability. Without a stable sequence of proposal versions, reasons for changes, staff analysis, meeting treatment, chair recommendation, Board action and implementation responsibility, an observer can know that a rule exists without being able to recover why this text rather than another became operative.

Fifth, online comments did not carry the same weight as comments offered during the meeting when consensus was determined. This was the most consequential diagnosis because it concerned not simply access but evidentiary influence. A person could participate through the stated online venue and still face a process in which an intervention made in the room mattered more at the decisive point.

Finally, the principles of openness, transparency and fairness were said to be followed implicitly but were not stated clearly enough to be broadly known or understood. This diagnosis was different in kind from the others. Stating principles can guide conduct and make deviations easier to identify, but principles do not execute themselves. “Fairness” cannot specify the treatment of a list objection unless the procedure defines the chain by which that objection is carried, answered and reflected in the outcome.

These were admissions in a proposal author’s drafting record. They are not a Board resolution, a judicial finding or proof that every described weakness produced harm in every case. Nor is the complete 2008 policy text available in the record examined here. The responsible claim is therefore limited: the 2010 proposer identified AFPUB-2008-GEN-001 as the affected process and described these as defects requiring replacement. The later 2011 implementation analysis independently restated the central objectives in a three-part form, giving the diagnosis unusual institutional weight without turning it into a finding of misconduct.

Two days after the March post, then PDP-MG chair Vincent Ngundi answered questions concerning the publication of agendas, meeting minutes and the boundaries of chair responsibility. His response treated agenda practice as something that could be stated explicitly, called the request for a minutes rule useful to the live proposal and said that the proposal created an opportunity to clarify the chairs’ terms of reference. He presented revision through a policy proposal as the path for improving the existing PDP.

Ngundi’s message matters because it shows procedural uncertainty being handled inside the process rather than denied. But it should not be inflated. It was one chair’s response, not collective ratification of the full defect inventory. It proves that agenda, minutes and chair-scope questions were live and considered susceptible to written repair. It does not prove what all participants believed, how a Board voted or what mandate anyone possessed.

From custom to written machinery

The selected institutional act occurred on 20 May 2010: publication of AFPUB-2010-GEN-004 as a replacement draft for the existing PDP. The once-public canonical page for GEN-004 was returning a not-found response when checked on 11 August 2026. That broken page is a provenance gap and should be stated as such. The identity and substance of the proposal remain supported by the March mailing-list draft and by the presentation delivered at AFRINIC-12 in June, but the missing canonical page cannot be described as if it had been retrieved live on the later cutoff date.

At AFRINIC-12, Moonesamy presented the replacement as a move from implicit practice to documented practice. The slides named openness, transparency and fairness and described participation through the RPD mailing list and the Public Policy Meeting. They set out a proposal path in which anyone could write a proposal, review would last at least four weeks, a Public Policy Meeting would consider the text, final review would follow and approval would complete the sequence.

The deck also presented emergency variance, appeal and chair recall. A changes slide said version 3 expressly required drafts to be posted to RPD and limited section 5.2 to discussion at a Public Policy Meeting. It recorded the movement of chair replacement away from Board discretion and toward the Working Group, constraints on Board approval, an appeal body and conditions for appeal, a recall committee appointed by the Board, a Board-requested variance and an implementation section.

Those changes redistributed real control inside the organisation. Requiring a draft to appear on the list reduced the possibility of a text materialising only in a meeting. Moving chair replacement toward the Working Group reduced an area of unilateral Board choice. An appeal route gave disagreement a named channel instead of forcing it into improvised persuasion. A recall process recognised that the conduct of a chair might itself need review. An implementation section acknowledged that adoption and execution are different stages.

The proposal was not blind to operational fragility. Its final risk slide identified unreachable chairs, an author unwilling to compromise, a disengaged community, bad implementation and politics. Those were risks named by the presentation, not evidence that each had occurred. Still, the list is revealing. It places human availability, incentives, participation, execution and institutional contest in the same field. A written procedure could constrain each of them, but no text could make them disappear.

The June discussion also demonstrated the value of forcing vague powers into language. On 1 June, mailing-list participant Walubengo J asked what conduct could justify a recall, who would replace a recalled chair, what counted as an emergency and who would authorise a variance. The quoted discussion included concern that the PDP-MG held too large a role. These questions do not prove capture or abuse. They prove that participants saw important ambiguities in the proposed safeguards themselves.

A mechanism may be named without being bounded. “Recall” is not yet a safeguard if the trigger, decision-maker and replacement process are unclear. “Emergency” is not yet a constraint if the condition, scope, duration and reviewer remain undefined. “Appeal” does not establish independent review merely by having a title. The questions on the list were therefore not peripheral objections to housekeeping. They went to the design of the control surface: who could interrupt the normal path, who would judge whether interruption was justified and how the process would continue afterward.

The strongest account of GEN-004 must nevertheless credit what it achieved. It exposed defects rather than concealing them. It kept drafting open, supplied a minimum review interval, linked the list to a public meeting, stabilised texts through stages, introduced final review, added dispute and recall machinery, constrained some Board discretion and generated a later implementation analysis. In a young, volunteer-dependent technical process, some informality may have been practical. Judgment by chairs is also unavoidable; neither written rules nor issue ledgers can mechanically calculate rough consensus.

That counterargument is substantial because this was self-correction, not institutional paralysis. The proper criticism is not that judgment existed or that every exercise of it was wrongful. It is that the replacement text did not directly make two of its own three central objectives reproducible. AFRINIC’s later record says so. The improvement was real, and the remaining hinge was real.

The chairs as a translation surface

The adopted process distributed action among proposal authors, mailing-list participants, meeting participants, chairs, AFRINIC staff and the Board. Yet distribution of participation did not mean distribution of decisive control. The chairs stood at the point where an extended, asynchronous record had to be compressed into a meeting account and then into an assessment of consensus.

The 2011 staff analysis listed chair tasks that made this position explicit. Chairs would summarise mailing-list discussion, chair face-to-face meetings, declare consensus, report, initiate Last Call, report to the Board and evaluate emergency variances. Each task is intelligible on its own. Combined, they put one role across collection, interpretation, procedure, decision framing, escalation and exception handling.

This does not make the chairs judges or regulators. They were administrators of a private policy process. But within that process their practical discretion was consequential. The person who summarises an objection decides, even without intending to, which details travel. The person controlling the meeting decides when discussion is complete and which questions receive time. The person declaring consensus transforms a heterogeneous record into a conclusion. The person initiating Last Call controls passage into the final review stage. The person evaluating a claimed emergency helps determine whether normal safeguards will hold.

When one role performs all those translations, openness at the input does not guarantee fidelity at the output. A list may contain a durable comment, but a meeting attendee may never hear it. Minutes may record that discussion occurred without linking the discussion to a stable issue identifier or exact proposal version. A final report may announce consensus without setting out unresolved objections and the evidence used to treat them. Nothing in this sequence requires bad faith. Compression itself creates selection.

The 2011 note’s treatment of online equality is therefore pivotal. It did not report that chairs had always favoured the room, and it cannot support that claim. It said the policy did not directly address equal weight and that the objective would depend on how chairs ran face-to-face deliberations. A stated right to comment existed beside an unwritten rule of effect. The procedural repair broadened visibility while leaving the most sensitive comparison—how remote and in-person contributions counted—inside conduct.

This is the difference between documenting discretion and eliminating it. A published chair report can reveal the result of judgment. A meeting minute can show that a subject was discussed. A proposal history can show that text changed. These records improve accountability, but unless they map each material issue through the process, they may not show why one objection altered the draft and another did not. Documentation can illuminate a discretionary hinge while leaving the hinge intact.

Nor should equality be reduced to counting comments. The record contains no defensible attendance denominator, no roll call, no remote-participant denominator, no vote tally and no reproducible consensus calculation for the June meeting. Mailing-list volume would not by itself establish support, and microphone sequence would not by itself establish legitimacy. The relevant design question is qualitative and traceable: did every material claim and objection receive identifiable treatment in the record on which the decision rested?

That standard preserves room for judgment while making judgment reviewable. A chair could explain that two objections were duplicates, that one depended on a superseded version, that another exposed an operational risk and that a final objection remained unresolved but did not defeat the narrow registry rationale. The point is not to replace facilitation with arithmetic. It is to prevent the passage from discussion to consensus from becoming an uninspectable act of compression.

Records are part of the decision, not decoration

The second objective left incompletely addressed—documentation of procedure, approval and implementation—was not merely a communications concern. In a versioned policy process, the record defines what later observers can verify. A rule cannot be evaluated only by its final prose; the chain of reasons, changes and institutional acts determines whether people can reconstruct how it came to apply.

The 2011 implementation analysis responded with a practical list. Version and event histories would identify the text and stage. Change logs with reasons would connect amendments to issues. Analyses and staff comments would disclose operational interpretation. Chair reports would carry discussion and consensus assessments forward. Meeting minutes would preserve deliberation. Published Board decisions would disclose the private corporate act that completed approval.

Together these materials could create an issue-to-decision chain, but only if connected. A pile of documents is not necessarily a record. A version history without stable reasons shows sequence but not causation. Minutes without issue identifiers show conversation but not disposition. A Board decision without the exact recommended text and unresolved objections shows approval but not the object approved. An implementation notice without responsibility and timing shows intention but not execution.

The current absence of the once-public canonical GEN-004 page illustrates the long horizon of this problem. Other official artifacts can preserve much of the substance, and the surviving mailing-list messages and presentation were still retrievable and matched the copies used for this analysis on the cutoff date. Yet broken canonical pages weaken the reader’s ability to move from an identifier to the authoritative record. Link durability is not glamorous, but it is part of institutional memory.

The record also says the Board had approved the new PDP and that it took effect during AFRINIC-13 on 11 November 2010. That chronology comes from the February 2011 implementation message. It does not reveal the Board’s reasoning, a vote or a public-law mandate. Approval was an act of a private corporate organ in the organisation’s internal sequence.

This boundary is especially important when discussing “implementation”. Applying a private process to the organisation’s own registry work may have contractual and operational consequences. It does not make the approval analogous to legislation. A complete archive can prove the sequence by which an internal rule was made. It cannot, merely through completeness, enlarge the institution’s authority over operators, assets, nations, users or people who were absent from the process.

Participation is evidence, not representation

GEN-004’s openness deserves respect on its own terms. Allowing anyone to propose text and maintaining public venues can expose technical facts, operational experience and objections that a closed drafting group would miss. The list can preserve asynchronous contributions across time and distance; the meeting can test claims through direct exchange. Used well, the two venues are complementary.

But participation is not representation. A mailing list is not a people, a Public Policy Meeting is not a legislature, and a self-selected set of contributors does not acquire sovereignty by being open to additional self-selected contributors. The word “community” may describe a field of people who engage with the process. It does not identify an electorate, enumerate represented principals or specify the authority delegated by each operator.

This is not a criticism of conversation. It is a limit on what conversation can authorise. A policy room can produce evidence and technical advice. It can help a registry protect uniqueness, registry accuracy, contactability, fraud control, transfer recording, security assertions, dispute metadata and operational continuity. Those functions are tied to the narrow work of maintaining a reliable registry. An open process can improve the quality of rules serving them.

Where a proposed rule goes beyond such an identifiable registry-function invariant, however, openness cannot supply the missing mandate. Binding effects must come from an applicable contract, competent public law or the voluntary consent of the affected principal—not from rhetorical expansion of “consensus”, “region”, “stewardship” or “fairness”. An internal procedure may order AFRINIC’s work. It may have consequences within an applicable agreement. Neither fact converts AFRINIC, its Board or its chairs into a legislature, regulator, court, police power or sovereign.

That distinction also disciplines the criticism of GEN-004. The problem is not that the process failed to become a better government. It should never be mistaken for government. The problem is that a private coordination system recognised defects in its own evidentiary and procedural chain, repaired some of them, and left important influence inside administrative discretion. Its proper measure is whether it protects the narrow registry function while preserving continuity, verifiability and exit for affected operators.

The sources supplied by Heng Lu, NRS, LARUS and BTW establish this institutional frame, not a rival vote inside AFRINIC’s process. The controlling principle is that attendance cannot be laundered into mandate and coordination cannot be inflated into public authority. The official AFRINIC record then establishes the words, dates, procedural roles and self-assessments to which that principle is applied. Keeping those functions separate avoids two errors at once: treating AFRINIC’s self-description as proof of legitimacy, or treating doctrinal criticism as if it could rewrite the historical record.

What the partial repair changed

GEN-004 materially altered the visibility of the process. A required list posting made the proposal text available before the meeting. A minimum review interval created time for asynchronous consideration. The staged path made it easier to identify when a proposal moved from discussion to meeting consideration, final review and approval. Appeal and recall mechanisms gave named form to disputes that previously lacked a stated route. Constraints on some Board choices narrowed internal discretion. Implementation received an express place in the lifecycle.

These changes lowered some risks. A proposal could be less easily detached from its public draft history. A person who disagreed with chair action could identify a process rather than improvise one. An absent chair need not leave the organisation with no recognised continuity mechanism. A version could be held stable long enough for people to discuss the same text. Staff could publish the sequence of events rather than leave later readers to assemble it from scattered messages.

The repair also made future diagnosis easier. The February 2011 implementation note was possible precisely because the organisation had named goals against which practice could be assessed. Institutional learning often proceeds in this way: a first text makes custom visible, its implementation exposes the remaining gaps, and subsequent operating procedures add connective tissue.

Yet the analysis’s three-goal test shows that connective tissue was doing work the policy itself had not completed. The dispute mechanism was explicit. Documentation was to be produced through a set of practices. Equal influence between venues rested on chair conduct. The replacement therefore moved from an under-specified process to a more legible hybrid of written rules and administrative judgment.

That hybrid may be functional. No serious institution can specify every contingency. The analytical question is which contingencies may safely remain discretionary. Choice of meeting order may require facilitation. Deciding whether a list objection was materially answered, whether an emergency justifies variance or whether an unresolved concern is compatible with consensus bears more directly on the rule’s legitimacy within the private process. The greater the consequence, the stronger the need for an explicit trigger, reason and review path.

One useful distinction is between discretion that fills detail and discretion that determines scope. A chair choosing how to group duplicate comments fills procedural detail. A chair deciding that a material remote objection need not affect the consensus account controls evidentiary scope. Staff choosing a page layout fills detail. Losing the authoritative version or publishing an approval without its reasons controls what outsiders can verify. A Board selecting dates may fill administration; using a variance without bounded grounds can alter the safeguards themselves.

GEN-004 reduced some scope-setting discretion but did not remove it. Its achievement should be measured as partial institutional learning: significant enough to deserve credit, incomplete enough to require a clearer architecture.

Operator impact without invented economics

The evidence does not show how many operators were affected by any procedural defect. It contains no price effects, quantified delays, deployment totals, loss estimates or market shares. It cannot support a claim that a particular chair action increased a particular operator’s costs. It does not establish fraud, corruption, censorship, conspiracy, confiscation or capture.

The economic mechanism can still be described cautiously. Registry rules can alter the conditions under which operators obtain, register, transfer, document or continue using number resources. Changes to those conditions can affect transaction work, deployment sequencing, compliance effort and continuity planning. If the policy-making chain compresses relevant evidence or cannot be reconstructed, the resulting rule can transmit an unexamined assumption into those operational conditions.

The effect need not begin with dramatic exclusion. It can begin with uncertainty: which text governs, how an exception is assessed, whether a remote objection was answered, when an implementation step becomes effective, or which actor has responsibility for correction. Operators respond to uncertainty by spending time, retaining options, delaying irreversible changes or accepting administrative risk. The sealed historical record does not measure these responses, so they remain a mechanism rather than a quantified finding.

The missing denominator is itself informative. Without reliable counts of participants, affected operators and represented principals, the process cannot plausibly infer a regional mandate from attendance. That does not invalidate technical input. It limits the conclusion that may be drawn from it. A strong argument by one operator can expose a registry risk; a hundred attendees cannot create sovereign authority merely through presence.

For that reason, the operator should ask two separate questions of every procedural reform. Did the reform make the registry’s internal decision more accurate and reviewable? And what source of authority makes the resulting rule applicable to this operator? GEN-004 addressed the first question imperfectly. It could not answer the second simply by improving participation.

What cannot be concluded

Several absences constrain the account. The complete text of AFPUB-2008-GEN-001 is not part of the evidence examined, so the older process’s defects should be described only as the 2010 proposer and 2011 implementation note described them. The exact first-submission message from December 2009 is not available here, so that date supplies context rather than the primary event. The canonical GEN-004 page and the current implemented-PDP page were not live at the 2026 cutoff, although surviving official artifacts preserve the relevant identity and substance.

There is no full transcript for the June 2010 meeting, no attendee or remote-participant denominator, no roll call, vote tally or reproducible consensus calculation. It is therefore impossible to quantify a meeting mandate, identify a regional electorate or establish how many comments were discounted. The presentation proves what the proposal claimed and what risks it named; it does not prove that each safeguard worked or each risk occurred.

The participant emails prove that named participants made particular observations or asked particular questions. They do not prove collective support, representativeness or motive. The February 2011 message proves the staff author’s implementation interpretation. It does not retroactively establish every fact about the meeting or show that equal weighting never happened in practice.

Nothing in the direct event record proves bad faith by any chair, Board member, employee or participant. Nothing ties the 2010 act to any current Board member, receiver, supporter, critic or officeholder. Present-day personalities are therefore outside the event. The continuing relevance is an abstract design test, not an allegation about current conduct.

These limits do not weaken the central finding. They sharpen it. GEN-004’s own record identified specified procedural defects. The replacement created genuine safeguards. AFRINIC’s later implementation analysis said that only the dispute mechanism was explicitly satisfied, while documentation required administrative work and venue parity remained dependent on chair practice. That narrow chain is enough to show the gap between an open input process and a reproducible decision.

A bounded alternative

A better design would keep the useful features of GEN-004 rather than discard them. Anyone could still draft. The public list and meeting could still collect different forms of knowledge. Chairs could still facilitate, synthesise and exercise judgment. The Board could still perform the limited corporate acts assigned to it. The change would be to surround each consequential judgment with a stable evidence path and a defined scope.

Every proposal version would have a durable identifier. Every material objection would receive an issue identifier and appear in a ledger that records the exact text it concerns, the evidence offered, the response, any textual change and its status at the meeting. The meeting agenda and chair summary would map back to that ledger. Minutes would identify not merely that a topic was discussed but what disposition was proposed. The final consensus explanation would name unresolved objections and explain why the narrow registry-function rationale did or did not answer them.

Chair absence and replacement would have automatic, time-bounded triggers. Emergency variance would require objective conditions, a written reason, a narrow scope, a duration and post-event review. Appeal and recall review would be separated enough from the conduct under review to test compliance rather than simply restate the original judgment. Implementation would name the responsible actor, authoritative text, effective date, dependent systems and rollback or correction path.

Most importantly, each proposal would identify the registry invariant it protects. Does it preserve uniqueness, improve registry accuracy, maintain contactability, control fraud, record a transfer, verify a security assertion, preserve dispute metadata or ensure operational continuity? If it does, the process can test whether the rule is proportionate to that function. If it reaches beyond those invariants into operator rights or obligations of a different kind, openness and internal consensus are insufficient. Applicability must come from contract, competent public law or voluntary affected-principal consent.

This design would not abolish politics, disagreement or error. It would make their route through the system inspectable. It would also reduce the temptation to treat the word “community” as both input and authority. The public process would remain a source of evidence. The registry would remain a private coordinator. Operators would be better able to verify which rule applies, why it exists and how to challenge procedural failure without conceding a mandate that the process never possessed.

The durable insight of GEN-004 is therefore neither triumph nor scandal. A private institution noticed that its process depended too heavily on implicit practice. It wrote more machinery, moved some control points and named routes for dispute. Then its own implementation analysis acknowledged that two central objectives still lived outside the text. That is a rare and useful record of institutional honesty. It shows how governance improves—and how easily procedural openness can be mistaken for authority unless the boundary is stated at every stage.