Summary

  • Draft 5 confined random review to Medium-and-above members, IPv6-only Large members and EU-AS members. Draft 6, dated 6 April 2018 and circulated in full on 10 April, replaced that category-limited class with a rule under which any member could be selected. AFRINIC staff asked whether this section 13.3.1 edit was the only change; an author expressly said it was. Co-chairs announced the update on 14 April.
  • The widened lottery disclosed no sampling algorithm, seed, cycle, frequency, sample size or public verification method. Membership-wide eligibility therefore did not mean that every member had been, or necessarily would be, reviewed in any stated period. Nor did it make random review equivalent to the draft’s separate selected and reported review lanes.
  • Everything consequential behind the entrance remained: evidence demands, staff findings, contact and correction, a three-month compliance or approved-transfer interval, contemplated recovery, database change, possible reallocation and an appeal phrase. The proposal remained Under Discussion. For an operator whose routing, customers and asset planning depended on resources recorded by a private bookkeeper, the stakes reached far beyond tidy records.
  • Removing categories repaired one distributional defect, not the mandate defect. AFRINIC could justify a narrow, contract-grounded check of identity, control, contact, uniqueness and record integrity. Randomness could not give it sovereign, regulatory, police, punitive, confiscatory or public-law adjudicative authority, because a private technical coordinator has none to give itself.

Summary

  • Draft 5 confined random review to Medium-and-above members, IPv6-only Large members and EU-AS members. Draft 6, dated 6 April 2018 and circulated in full on 10 April, replaced that category-limited class with a rule under which any member could be selected. AFRINIC staff asked whether this section 13.3.1 edit was the only change; an author expressly said it was. Co-chairs announced the update on 14 April.
  • The widened lottery disclosed no sampling algorithm, seed, cycle, frequency, sample size or public verification method. Membership-wide eligibility therefore did not mean that every member had been, or necessarily would be, reviewed in any stated period. Nor did it make random review equivalent to the draft’s separate selected and reported review lanes.
  • Everything consequential behind the entrance remained: evidence demands, staff findings, contact and correction, a three-month compliance or approved-transfer interval, contemplated recovery, database change, possible reallocation and an appeal phrase. The proposal remained Under Discussion. For an operator whose routing, customers and asset planning depended on resources recorded by a private bookkeeper, the stakes reached far beyond tidy records.
  • Removing categories repaired one distributional defect, not the mandate defect. AFRINIC could justify a narrow, contract-grounded check of identity, control, contact, uniqueness and record integrity. Randomness could not give it sovereign, regulatory, police, punitive, confiscatory or public-law adjudicative authority, because a private technical coordinator has none to give itself.

L3 — One deleted category line widened the lottery

The most revealing feature of Draft 6 is how little its authors said they had changed, and how much institutional machinery that small edit left standing. The instrument identified itself as AFPUB-2016-GEN-001-DRAFT06, Version 6.0, “Internet Number Resources Review by AFRINIC”. It carried the date 6 April 2018. Its complete text reached the RPD mailing list on 10 April. Four days later, on 14 April, the policy working group’s co-chairs announced receipt of the updated draft and asked participants to review and comment.

Those are three distinct procedural facts: the date on the instrument, the date of full circulation by the authors and the date of the co-chair notice. None should be collapsed into another.

The change had been foreshadowed on 31 March. The authors told the list that the working group had accepted the removal of membership categories from section 13.3.1 and that the next version would reflect it. When Draft 6 appeared, its revision history likewise described the deletion of categorisation among members in the random class. An AFRINIC staff member then put the version-control question directly: was that the only change from Draft 5? A proposal author answered yes. This exchange matters because it narrows the legitimate historical claim.

One need not infer a sweeping rewrite from a new version number, and one must not attribute other changes to Draft 6. The authors’ own account was that the random-class edit was the sole delta.

Draft 5 had named the categories from which a random review could be drawn. It referred to Medium and above, IPv6-only Large and EU-AS members. Draft 6 removed that classification and instead said that any member could be chosen at random from the membership. The immediate effect was enlargement of the eligible pool. A member outside the earlier named categories was no longer excluded merely by category. Conversely, the formerly named groups were no longer singled out in the text as the only candidates for random selection.

That is a real equality repair. Category-limited random selection creates an obvious distributional question: why should those sizes or resource types carry the chance of review while other members do not? Eliminating the categories gives the lottery a formally inclusive starting population. If the legitimate purpose is to test the accuracy of a registry’s records across its membership, drawing from the full set may also produce a broader picture than repeatedly looking at a few specified classes.

A private recordkeeper has sensible reasons to discover stale contact details, disputed control, duplicate claims or other errors that could undermine uniqueness and continuity. An inclusive sample can, in principle, help find those defects without presuming that a named category is especially suspect.

But “any member may be selected” is not the same proposition as “every member will be reviewed”. Draft 6 did not prescribe a universal inspection schedule. It did not say that the whole membership would pass through review within a year, within a cycle or at all. It set no cycle. It stated no frequency, cohort size or total sample. It did not disclose a method by which selection would be generated, a seed by which a draw could later be reproduced, or a public verification procedure through which members could know that the rule had been followed.

Membership-wide eligibility enlarged the universe of possible selections; it did not establish the number or timing of actual selections.

This distinction is central to judging both fairness and risk. A rule can be universal in eligibility yet uneven in operation. That observation does not support an allegation that Draft 6’s lottery was manipulated or biased. The published record does not show an operational algorithm at all, still less the operation of a corrupt one. The sound conclusion is narrower: the text did not make randomness auditable. Without a stated method, an outsider could not distinguish a properly generated choice from an unexplained one. Without a frequency and sample size, a member could not estimate ordinary exposure.

Without a cycle or exclusion history, the membership could not see whether repeated selection was possible, whether recently reviewed portfolios were removed from the pool, or whether the system distributed burden over time.

Random review was also only one entrance into the larger structure. Draft 6 carried forward a selected-review lane. Under that lane, a review could follow an internal report or the absence of contact. That is not random selection: it begins with an internal signal or a contactability problem. It also retained a reported-review lane, available after a member’s own request or after a community complaint supported by evidence and evaluated by AFRINIC staff. That is not random selection either: it begins with a request or a report that must enter an evidential assessment.

Keeping these lanes separate prevents several false conclusions. A request by a member to review its own position cannot explain the fairness of the membership lottery. An internal report is not a random number. A community complaint is not a random draw. Conversely, a transparent lottery would not by itself settle whether an internal report was reliable or a complaint sufficiently grounded. Each trigger needs safeguards appropriate to its source. A chance selection needs reproducibility and even distribution. A selected review needs a clear definition of the internal report and a fair way to resolve contact failure.

A reported review needs a defensible complaint threshold, protection against misuse and a way to test the evidence without treating allegation as fact.

The reported lane contained a sentence concerning the same resource portfolio within twenty-four months. Its location matters. The sentence remained within the reported class; the text did not express it as a universal protection against random or selected review. It is therefore inaccurate to read it as a blanket two-year limit on all forms of scrutiny. At most, on the face of the published proposal, it qualified repeated reported review of the same portfolio. The absence of an equivalent express constraint for random and selected review was one of the reasons that the widened random pool still lacked a measurable burden boundary.

Draft 6 also allowed AFRINIC staff, in handling a complaint, to require evidence in the formal form of a sworn affidavit or a declaration before a Commissioner of Oath. Formality can deter casual accusation and attach consequences to dishonest evidence. It can improve evidential discipline. Yet the fact that a statement is sworn does not prove its contents true. A formally presented complaint remains a claim requiring examination, disclosure of the material case to the affected member, and a reasoned conclusion. Ceremony strengthens accountability for the assertion; it does not transform assertion into adjudicated fact.

Beyond the entry lanes lay the architecture that Draft 6 did not alter. The text retained policy or agreement breach as grounds for a resource-recovery process. It contemplated attempted contact and an opportunity for correction. If the matter remained unresolved, it provided a three-month public interval during which the member could seek compliance or an AFRINIC-approved transfer. After that interval, it contemplated recovery, an update to the former holder’s records and possible reallocation under existing policies.

Thus the sole equality edit at the front of the random lane still fed into a sequence capable, on its own terms, of producing serious registry consequences.

The same was true of appeal language. Draft 6 used the phrase “within the four weeks”, referred to arbitration under Mauritian civil procedure, and allowed AFRINIC, on request, to suggest a pool of knowledgeable volunteer arbitrators. Yet the text did not clearly anchor when that four-week period began. Nor did it state that an appeal would automatically stay publication, recovery, database modification or reallocation. Those are not cosmetic omissions. A review right that arrives after an irreversible operational step is not equivalent to one that preserves the position while a neutral decision-maker examines the dispute.

Draft 6 also retained annual meaningful reporting of review activities, subject to applicable laws and regulations. Aggregate reporting can be useful. It can show the number and type of reviews, their duration, broad outcomes, error rates and whether burdens fall disproportionately. It may expose process drift without exposing a member’s sensitive material. But an undertaking to report annually is not permission to disclose confidential case details. Nor does a reporting commitment substitute for advance rules governing custody, access, purpose, retention and liability for evidence gathered during a review.

The instrument labelled itself Under Discussion. The co-chairs’ 14 April message sought further review and comment. The sealed record at this publication stage does not establish adoption, ratification, implementation or application to any member or resource. That status is indispensable to the analysis. The proposal is important because of the authority it claimed on paper and the questions participants immediately continued to raise, not because this record proves that the machinery was activated. Treating circulation as enforcement would manufacture a fact.

Treating an Under Discussion draft as harmless merely because it was not yet operative would miss the point of scrutinising institutional design before consequences occur.

The April responses show why the one-line edit did not close debate. Participants raised financial loss and voluntary leasing; asked for the random algorithm and timeframe; questioned compensation; and pressed on confidentiality, non-disclosure penalties, undefined procedures, audit cost and neutral external appeal. On 29 April, a participant said that standing objections had neither been withdrawn nor addressed and that significant changes remained necessary. These contributions prove that the concerns were voiced and persisted after Draft 6.

They do not prove that a confidentiality breach, financial loss, victimisation or unlawful act actually occurred. They do not prove bad faith by the authors or AFRINIC. And a participant’s view of consensus is not itself a formal adjudication of the proposal’s final status.

Seen in sequence, the historical point is restrained but powerful. Draft 1 had begun the proposal series on 18 May 2016. By the 6 April 2018 date of Draft 6, the text had travelled through six versions over nearly twenty-three months. The sixth revision removed a categorical inequality. Within days, however, questions about the lottery’s method and timeframe and about compensation, confidentiality, cost and independent appeal remained active. The duration alone does not prove invalidity; difficult policies can take time.

What it shows, together with the narrowness of the sole confirmed revision and the continuing objections, is that the institutional boundary had not been resolved merely by broadening the pool.

The redline therefore has two faces. On one face, it is a worthwhile correction: every member entered the same possible random class. On the other, it illuminates everything that did not change. It says who could be drawn but not how the draw would be made. It widens exposure but does not define the information demand. It equalises a chance of entry but does not allocate cost, guarantee confidential handling, compensate wrongful harm or preserve continuity during appeal.

And because the authors confirmed that section 13.3.1’s category removal was the only change, the carried-forward review-to-recovery chain cannot be treated as if Draft 6 newly repaired its authority foundations. The lottery became broader. The mandate did not become stronger.