Summary

  • AFPUB-2021-GEN-002-DRAFT03, submitted by Jordi Palet Martinez on 18 May 2022, proposed an integrated rewrite of AFRINIC’s Policy Development Process rather than a change to a single number-resource rule.
  • The proposal answered genuine procedural problems: stale versions, late impact analysis, compressed meeting debate, uncertain Last Call practice, appeals colliding with expiry, online-access constraints and competing texts.
  • Its machinery nevertheless allocated substantial private discretion among the author, participants, chairs, AFRINIC staff, Board-appointed appeal bodies and the Board, while corporate members and affected operators occupied different and sometimes indirect positions.
  • The legal relationship between the wider policy community and AFRINIC’s bylaws was disputed at the time. The author and AFRINIC counsel took opposing positions; neither position is treated here as a legal judgment.
  • A sounder model would preserve useful clocks and written reasons but attach every procedural power to a narrow technical purpose, an identifiable principal, an evidence record, independent review, continuity safeguards and a replayable rule-change receipt.

Seventeen pages aimed at the machine

The institutional act that matters occurred on 18 May 2022. On that date Jordi Palet Martinez submitted AFPUB-2021-GEN-002-DRAFT03, the third version of a proposal to revise AFRINIC’s Policy Development Process. Its target was not a particular allocation threshold, transfer condition or routing practice. It was the procedure that could make all such policies later. In seventeen pages, the draft tried to define the path from a proposal’s arrival to its expiry, from discussion to rough consensus, from Last Call to Board ratification, and from appeal or emergency back into ordinary process.

Timing gave the submission an institutional weight that the text alone cannot convey. AFRINIC-35 would open on 1 June, exactly two weeks later. Before Draft 3 arrived, AFRINIC’s Governance Committee had publicly identified a separate corporate risk: open Board seats and the sequencing of elections could leave the Board unable to raise ordinary quorum. The published bylaws required at least five directors for that quorum. This was a concern about a possible loss of normal governing capacity, not proof that AFRINIC lacked a Board on either 18 May or 1 June.

Nor does the proximity prove that the proposal was caused by the succession problem or could solve it. Yet the setting matters. A private registry was considering a more elaborate policy-making machine while publicly acknowledging that the support under its corporate decision-making could become incomplete.

That juxtaposition is the article’s starting point, not a claim of causation. Draft 3 did not create a Board vacancy, remove a director, cause an outage or decide a court case. It did, however, prescribe several points at which the Board would enter the policy process. It therefore asked readers to evaluate a procedural architecture whose final and exceptional gates depended on a corporate organ already surrounded by a documented capacity concern. The proposal was meta-governance under stress: an attempt to make policy formation more regular at the very moment when institutional continuity could not be assumed as mere background.

The official revision history sharpens the boundary. Draft 1 was submitted on 10 October 2021, Draft 2 on 9 November, and Draft 3 on 18 May 2022. AFRINIC-34’s record says Draft 2 did not achieve consensus on 18 November 2021 because valid objections and concerns remained, so it returned to the mailing list. The revision note for Draft 3 specifically identified updates related to expiry of a Policy Proposal Version, or PPV. These facts establish a proposal moving through versions and a meeting-stage failure to secure consensus. They do not establish that Draft 3 cured every earlier objection.

The distinction is essential because procedural history can be flattened too easily. Submission is not discussion. Discussion is not a chair’s consensus call. A consensus call is not Last Call. Last Call is not Board ratification. Ratification is not implementation, and implementation is not proof that a mechanism was invoked against anyone. The official proposal page records later events, but they are not the act examined here. The centre of gravity remains 18 May: the submission of a proposed integrated design, followed by the 1 June meeting only insofar as it explains what that design meant to contemporaneous actors.

A repair case that deserves to be heard

The strongest argument for Draft 3 begins with the inadequacy of improvisation. AFRINIC’s own published record contained recognisable process failures. Proposals could become dormant. Important analysis could arrive late. A short public-policy meeting might be asked to resolve questions that had not been adequately developed on the mailing list. Appeals could outlast the version they challenged. Last Call practice could be unclear. Different proposals could contain competing text. Pandemic-era constraints made participation dependent on whether online and hybrid formats were treated as real parts of the process or as lesser substitutes.

These are not cosmetic complaints. A number registry’s policy process affects the conditions under which records and resource-related decisions are coordinated. If participants do not know which version is live, when evidence will appear, how long objections can remain unanswered or which clock an appeal stops, the result is not simply messy minutes. It is uncertainty about the procedure itself. That uncertainty rewards insiders who can follow every thread, encourages decisive action at compressed meetings and leaves affected operators unable to predict when a proposal may change the environment around them.

Draft 3 tried to place those uncertainties on a common track. It enumerated twelve problem areas, covering the definition of consensus, the composition of the Policy Development Working Group, operation during Covid-era conditions, inclusion of meeting and list discussion, impact analysis, Last Call, expiry, appeals, timing, online sessions, Board functions, moderation and conflict resolution. The breadth could look overambitious, but it also expressed a coherent premise: these failures interacted. An expiry rule could not be fixed without deciding what an appeal did to time.

A meeting requirement could not be fixed without deciding whether online sessions counted. A consensus definition could not be fixed without assigning someone to classify objections. A Board gate could not be fixed without saying what the Board received and what it could return.

There is real institutional value in an early impact analysis, a published explanation for delay, a fixed Last Call, reasons for a failed proposal and an explicit path back to discussion. Regular clocks can limit selective urgency. Written reasons can expose whether a chair addressed a material technical concern or simply pronounced a result. Online participation can reduce the accident of geography and travel. Expiry can prevent abandoned text from remaining indefinitely actionable. An appeal pause can stop a version from dying merely because a challenge took time to decide.

A return route can prevent ratification from being treated as an unexplained yes-or-no ritual.

The benign case is therefore not that every clause was correct. It is that private coordination becomes safer when decisions are versioned, timed, explained and capable of correction. AFRINIC-34 had already supplied the immediate warning: Draft 2 returned to the list because objections remained. Draft 3’s answer was to join the stages into a more explicit machine. A fair assessment must grant that the machine addressed observed weaknesses rather than dismissing every procedural refinement as a bid for power.

Who entered the working group

The first allocation concerned participation. Draft 3 proposed that any real person could take part in the PDWG. AFRINIC could formally verify identity when necessary, with personal-data protection attached to that possibility. This appears inclusive at one level and restrictive at another. It opened the process beyond corporate membership, but it reserved an institutional function for deciding when a participant must prove personhood and how that proof would be handled.

Nothing in the record supplied here establishes that AFRINIC used this proposed language to investigate or verify any named person. It was a design power, not evidence of an exercise. Still, the clause mattered because identity is not a neutral administrative detail in an open technical forum. A real-person rule may answer concerns about impersonation, duplicated personas or unverifiable interventions. But identity verification can also chill participation, expose individuals working under constrained conditions or allow administrative judgments to affect whose contribution remains credible.

A well-bounded mechanism would need a trigger, a minimum data requirement, a retention rule, a means of challenge and a clear separation between identity assurance and viewpoint assessment.

The working group itself also occupied an ambiguous institutional position. It was broader than AFRINIC’s corporate membership and included people who might bring technical expertise, local knowledge, operator experience or objections. Those contributions are valuable. They do not, by themselves, establish that participants were authorised by every affected network, resource holder or customer. A mailing-list subscriber may be informed without being a representative. A meeting attendee may identify a serious flaw without holding a mandate to bind those absent.

Participation improves evidence; it does not answer the separate question of who has authorised a rule-maker.

That distinction becomes more important once the procedure gives participation an apparently constitutional vocabulary. Calling an open set of contributors “the community” can make a useful forum sound like a political body. But the label does not provide an audited constituency, a delegation instrument or a denominator of represented operators. The available evidence does not contain such a denominator. Draft 3 could specify who was eligible to enter the process. It could not, by eligibility alone, turn entry into authority over everyone downstream.

Two channels, several clocks

Draft 3 treated meetings and the Resource Policy Discussion list as joined but distinct channels. It proposed two normal Public Policy Meetings each year, which could be online, in person or hybrid, and allowed additional online PPMs. Other online input sessions could occur, but they would not count toward consensus. That distinction attempted to preserve a formal decision pathway while admitting that useful discussion could happen elsewhere.

For a proposal never presented at a PPM, the draft required at least eight weeks of list discussion and a meeting presentation before consensus could be determined. A later version could avoid repeat presentation if the chairs concluded that mailing-list discussion had resolved the objections raised earlier. This was a practical answer to duplication: not every revision needs a full ceremonial replay. Yet it placed a substantive judgment inside the waiver. Whether earlier objections were “resolved” would decide whether another meeting was required, and the chairs would make that call.

The text also gave the chairs up to two weeks after the list and meeting conditions were satisfied to determine rough consensus. PPM minutes were due within two weeks, and a PPV that failed to reach consensus required reasons. Those deadlines formed a useful audit sequence. Participants could know when the decisional window closed and, at least in principle, why the process did not advance.

Alongside discussion ran the impact-analysis clock. AFRINIC would have four weeks after submission of each PPV to publish an analysis and, in any event, would have to publish it at least one week before the PPM. If it could not meet the timetable, it would have to justify the delay and publish at least a draft analysis. This clause addressed a real asymmetry. Proposal authors can circulate text before the registry has explained implementation, legal, operational or financial consequences. If the analysis arrives at the last moment, participants may have nominal access but insufficient time to assess it.

Early analysis is therefore one of Draft 3’s most defensible devices. Yet the design made AFRINIC both an affected institution and the producer of the official impact account. Staff knowledge is indispensable: the registry understands its systems, operational constraints and internal dependencies. But an impact assessment is evidence supplied by an interested actor, not independent validation of the proposal’s legitimacy or efficacy. Its scope, assumptions, omissions and timing must remain contestable. A deadline makes the evidence visible; it does not make it conclusive.

Expiry introduced another clock. Draft 3 proposed that a PPV expire after six months unless the chairs had submitted it to the Board for ratification. A new version reset the clock. An appeal paused the expiry counter, and a later update to an expired PPV could retain the same proposal identifier. These choices tried to balance freshness with continuity. Text would not linger indefinitely, but a proposal’s history would not disappear merely because one version timed out.

The appeal rule shows why clocks cannot be assessed separately. Draft 3 retained a threshold of three supporters for an appeal and five for a chair recall. It said appeals and recalls would not alter PDP timing, while also requiring the Board to hold ratification when an appeal remained pending. At the same time, an appeal paused PPV expiry. These are distinct effects: the broader procedure could continue to have timing rules, the challenged version would not expire solely because the appeal consumed time, and the Board could not complete ratification while the appeal was unresolved.

The architecture sought to prevent either a challenge or a clock from silently destroying the other.

Consensus as classification, not counting

Draft 3’s proposed rough-consensus definition rejected voting and headcounts. It required critical technical objections to be addressed. It also allowed chairs to discount repeated support for an objection they considered technically refuted. This model reflects a familiar engineering intuition: ten repetitions of a weak claim do not outweigh one unresolved technical failure, while ten endorsements do not repair a broken design.

That is the strongest reason to use rough consensus. It focuses attention on the quality of objections rather than the popularity of camps. A routing or registry-coordination proposal should not succeed merely because supporters mobilise more voices. If a single participant identifies a credible threat to uniqueness, accuracy, proof of control, security or continuity, the process should address the substance.

But rejecting headcounts does not remove power. It relocates power into classification. Chairs must decide which objection is critical, whether it is technical, whether it has been answered and whether renewed support adds information or merely repeats a refuted point. Those decisions may be careful and expert. They are still decisions. A declaration of rough consensus can conceal the operative reasoning unless the objection ledger and chair treatment are explicit.

Draft 3’s requirement to give reasons when a PPV failed was helpful but incomplete as a general control. The same transparency is needed when a proposal succeeds. Readers need to see the strongest objections, the evidence offered in response, the person or role that classified them, and the explanation for treating them as resolved, out of scope, editorial or still material. Otherwise “rough consensus” becomes a result word whose technical pedigree shields an unreviewable exercise of discretion.

This does not require turning the process into a vote. Voting would answer a different question and could create its own false appearance of representation. Nor does it require endless accommodation of objections that have already received a reasoned technical answer. It requires replayability: a later reader should be able to reconstruct how the forum moved from competing claims to a chair’s conclusion without relying on trust in the conclusion itself.

Last Call and the difference between correction and reopening

After a consensus determination, Draft 3 fixed Last Call at two weeks and gave the chairs one additional week to confirm whether consensus remained. It admitted editorial comments and exceptionally permitted substantiated new objections. If an editorial modification was made, Last Call would restart.

This design recognised that final review has two conflicting purposes. It must catch mistakes in the text that supposedly reflects agreement, but it must not become an unlimited second beginning whenever a participant dislikes the outcome. By distinguishing editorial corrections from exceptional new substantive objections, Draft 3 tried to keep the final stage bounded.

The difficult word is “editorial.” A wording change can look minor while changing who falls within a rule, when a condition is met or what evidence is required. Restarting Last Call after editorial modification was therefore a prudent safeguard. It acknowledged that even non-substantive correction should be exposed to a fresh, fixed review period rather than silently incorporated. The same logic, however, calls for a public redline and a reason for the classification. Without those, readers cannot tell whether the restarted period concerns punctuation, a repaired ambiguity or an altered obligation.

Last Call also illustrates why procedural sequence must remain precise. A successful chair call did not itself enact policy. The proposal still had to complete Last Call, preserve consensus, reach the Board and obtain ratification under the proposed design. Any account that collapses those stages gives consensus more force than the draft itself assigned to it.

The Board gate and the emergency branch

After consensus, Draft 3 required the chairs to send the proposal to the Board with a short report of the discussion. The Board would ratify it or return it to the PDWG with reasons and possible alternatives, then communicate its decision. The path was deliberately two-way. A returned proposal would go back to the forum rather than vanish behind a corporate refusal.

That structure is more accountable than an unexplained veto, but it exposes the relation between two constituencies. PDWG participation was open to real persons. AFRINIC’s Board belonged to a private member-based company. Corporate members and wider participants were not identical groups, and neither could simply be treated as the other’s agent. The Board gate therefore did more than perform administrative completion. It was the point at which an open technical process met private corporate responsibility.

Draft 3 also proposed that, in a duly justified emergency, the Board could define temporary policy changes until the next PPM. The change would enter the PDP. If the future process did not reach consensus, further enforcement or implementation would stop, while prior actions would remain valid. A separate provision for varying normal timing required clear and sufficient justification, four weeks of discussion, a one-week Last Call and later presentation at a PPM after consensus.

These provisions responded to a real concern: a registry may face a time-sensitive coordination problem that does not wait for the ordinary calendar. But emergency power is where procedural convenience comes closest to substantive authority. The proposed branch placed initiation in the Board, allowed action before the ordinary process had run its course, and protected prior actions even if later consensus failed. Its legitimacy therefore depended on a narrow definition of emergency, a tight technical scope, a short duration, published reasons, continuity protection and a return to ordinary procedure.

The available record does not show that the Board used Draft 3’s temporary-policy authority. Draft 3 remained a proposal at the act examined here. The clause is evidence of a proposed allocation of control, not evidence of invocation, enforcement or harm.