Summary

  • AFRINIC’s reconstituted Governance Committee has five voting members: three elected by the Membership and two appointed by the Board.
  • Formal advice must pass three tests at once: at least 65% approval, no more than 20% dissent, and no more than 30% abstention or failure to vote.
  • A conflict recusal is removed from the eligible-voter denominator, while an absent eligible member remains in it. With one recusal, a 3–1 vote has 75% approval but 25% dissent, so it cannot become formal advice.
  • Any future published advice would be easier to audit if it carried a small calculation table showing the denominator, recusals, absences, vote totals and the result of each test.

Four yeses are the normal five-person rule

AFRINIC announced the constitution of its Governance Committee on 17 August 2026. The voting room contains five seats. Daniel Khauka Nanghaka, Afaf El Maayati and Maud Adjeley Ashong Elliot were elected by the Membership; Phumzile Madonsela and Eddy Lareine were appointed by the Board. The Board liaison, Legal Adviser and Secretariat participants do not vote.

That composition has already attracted the obvious governance questions: who selected each member, when the terms expire, and how non-binding advice returns to the institution that requested it. A less visible control sits inside the committee’s Terms of Reference. Five equal votes do not become formal advice merely because a familiar majority or even a familiar supermajority supports the proposal.

The committee should prefer perfect consensus. If it votes, three conditions must all hold. Approval must be at least 65% of eligible voting members. Dissent must be no more than 20%. Abstention or failure to vote must be no more than 30%. The word “and” does the institutional work. Passing one or two of the conditions cannot compensate for failing the third.

AFRINIC supplies the base-case arithmetic. With five eligible members, formal advice needs at least four YES votes, no more than one NO, and no more than one abstention or failure to vote. Three distributions therefore pass: 5–0–0, 4–1–0 and 4–0–1, in the order YES–NO–abstention or failure. A 3–2 result is not close formal advice. It misses the approval floor and breaches at least one ceiling.

This is more exacting than the phrase “65% approval” sounds on its own. In a five-person body, 65% rounds the required support up to four. The dissent and non-voting ceilings then describe which fifth vote can sit beside those four. The committee is not using a single hurdle; it is using a three-part classification rule.

A recusal changes more than the headcount

The Terms of Reference require a member with an actual or reasonably perceived conflict of interest to recuse from the affected vote. A recused member is not counted among eligible voting members. By contrast, an absent eligible member remains inside the denominator. Those two empty chairs may look identical on a meeting screen, but the rule treats them differently.

Now remove one member through a recusal. Four eligible voters remain. Three YES votes equal 75%, safely above the 65% approval floor. One NO vote, however, equals 25%, above the 20% dissent ceiling. The motion therefore fails as formal advice even though three of four eligible voters support it and nobody abstains.

The permissible four-person distributions expose the point. A 4–0–0 vote passes. A 3–0–1 vote also passes: 75% approve, nobody dissents and 25% abstain or fail to vote. A 3–1–0 vote does not pass. The same three YES votes have not changed; the classification changes because the remaining vote moves from the 30% channel into the tighter 20% channel.

This is editorial arithmetic applied to AFRINIC’s published rule. It is not a report that anyone has recused, voted 3–1, dissented or blocked advice. No current dispute can be inferred from the calculation. The value of the exercise is prospective: the new five-person committee now has a rule whose outcome can turn on which denominator applies and how the fourth vote is recorded.

There is a separate membership floor. Section 7.8 says that if the committee has fewer than four voting members, it may give no formal advice until its voting membership returns to at least four. That is not the same as saying a recusal automatically invokes the four-member floor. A person may remain a voting member of the committee while being ineligible for one conflicted vote. Conflating membership with vote-specific eligibility would change the published rule.

Absence is not a silent recusal

The denominator also gives absence a precise effect. If all five members are eligible and one fails to vote, that non-vote remains in the denominator and occupies the abstention-or-failure channel. The other four must all vote YES for formal advice to pass. Four yeses produce 80% approval; the single failure to vote is 20%, below the 30% ceiling; there is no dissent.

If one of the four participating members votes NO while the fifth is absent, the result becomes 3–1–1. Approval falls to 60%, dissent is 20% and absence is 20%. Two ceilings are respected, but the approval floor fails. Again, the rule does not ask which condition feels most important. It asks whether every condition is true.

This design separates several kinds of institutional information. A recusal signals that a conflict rule has removed someone from a particular decision. An absence records that an otherwise eligible member did not participate. A NO records substantive dissent. An abstention records presence without support or opposition. Combining them into “votes cast” would conceal exactly the distinctions the percentages were designed to preserve.

The vote method may be a secret ballot, roll call, show of hands or an electronic method. The chair chooses the method, although an eligible voting member may insist on a stronger one. That flexibility does not make the threshold unknowable. Individual choices may remain secret where the selected method requires it, while aggregate YES, NO and abstention-or-failure totals still establish whether formal advice passed.

Formal advice is an output, not a command

The importance of the calculation should not be exaggerated into a larger authority. GovCom may advise the Board, the Membership or the community on matters within its scope. Its advice is expressly non-binding. If the Board chooses not to follow formal advice directed to it, the Board must publish reasons. Membership and community processes retain their own decision routes.

The vote therefore controls the label and institutional provenance of a recommendation. It answers whether a text can leave the committee as formal Governance Committee advice. It does not allocate IPv4 space, revoke a member’s holdings, direct a route announcement, decide litigation or turn AFRINIC into a sovereign authority over working networks. Those effects require their own lawful actors and processes.

That boundary makes the arithmetic more useful, not less. Because recipients remain free to accept, reject or process the advice elsewhere, readers need to distinguish three moments: the committee’s threshold decision, the advice and reasons it publishes, and the recipient’s later response. A tally cannot prove implementation; a response cannot retroactively prove that the committee cleared its own voting rule.

The Terms of Reference already require minutes and prompt publication of formal advice and its reasons. They also require the Board to publish reasons when it declines Board-directed advice. A compact voting calculation would sit between those existing records. It would describe how an item acquired formal status without pretending that formal status settled the underlying issue.

Publish the three tests separately

The useful evidence object is not another general promise of transparency. It is a small, repeatable table attached to each published formal advice item. The table should identify the number of voting members at the time, the number eligible for this decision, the number recused, the number absent, and the aggregate YES, NO and abstention-or-failure totals.

Then it should show three rows. The approval row gives YES divided by eligible voters and marks whether the result is at least 65%. The dissent row gives NO divided by eligible voters and marks whether it is no more than 20%. The abstention/failure row gives the residual category divided by eligible voters and marks whether it is no more than 30%. The formal-advice outcome is PASS only when all three rows pass.

This proposal does not require publication of a conflicted member’s private information. A record can identify that a recusal occurred and state the rule applied without narrating personal details beyond what a lawful conflict process requires. Nor does it require exposing individual votes from a secret ballot. The denominator, aggregate categories and calculation are enough to test the outcome.

The table would also protect the committee from misleading shorthand. “Approved by 75%” sounds decisive until the reader learns that the remaining 25% was dissent rather than abstention. “Three votes in favour” sounds incomplete until the reader knows whether the denominator was four or five. A three-line calculation turns both impressions into a reproducible result.

No current formal-advice item is assessed here. AFRINIC’s public meeting-report page describes 15 reports from 2017 through 2022, and the visible register contains historical requests through 2020. Those pages establish the state of the reviewed public surfaces, not the absence of current meetings, requests, votes or internal work. The test should be applied when a current item is actually published, not invented in advance.

What the sources establish

The 17 August announcement and current committee page establish the five voting members, their three-to-two selection split and the non-voting roles. The Terms of Reference establish the three simultaneous thresholds, recusal and absence treatment, the separate minimum-membership rule, publication duties and non-binding status. The meetings and register pages establish only what their public indexes displayed at research time. The 4–0–0, 3–0–1 and 3–1–0 classifications for four eligible voters are calculations from those rules, not reported events.

Sources