Summary

  • AFRINIC’s 1 September publication presents Version 3 of the RIR Governance Document as a recommended final draft, not an adopted rule or a live status-change case.
  • The draft calls the registry system bottom-up, but new Article 2.7 says it neither prevents nor obliges an RIR or ICANN to consult its community while assessing Recognition or Derecognition.
  • Recognition still requires broad resource-holder and community support; Derecognition still requires attributed proposals, responses, assessments and a two-thirds RIR threshold.
  • The proportionate control is not a referendum. It is a public receipt showing who chose whether to consult, which proposal version and community were in scope, what evidence was used and why.

Two sentences define the tension

AFRINIC’s 1 September notice does something useful before it makes any institutional claim: it identifies the material as the NRO Number Council’s recommended draft and links the text, redline and rationale. The notice does not say that the document has entered into force. It announces a finished recommendation and says that further opportunities for input will be announced by the NRO Executive Council.

The draft’s preamble describes the Internet Numbers Registry System as decentralised and coordinated, based on open and bottom-up processes. Its stated goals include preserving the bottom-up, multistakeholder model. Those phrases are not decorative. They tell members, resource holders and network operators where legitimacy is supposed to come from when the recognised regional institution changes.

Article 2.7 then supplies a precise limit. Nothing in the document, it says, either prevents or obliges the RIRs or ICANN to conduct consultations with their respective communities when assessing a proposal for Recognition or Derecognition.

The rationale removes any doubt that this was an accidental gap. Article 2.7 is new in Version 3. It was added after comments said it was unclear when stakeholders would be consulted. The answer was not a consultation calendar. It was an express preservation of discretion.

That can be defensible. “Bottom-up” does not mean that every affected person acquires a vote over every corporate, contractual or coordination decision. A stakeholder bears consequences; a principal holds a decision right. Consultation can improve the evidence available to the principal without becoming the source of the principal’s authority.

The difficult question is therefore not whether consultation must always occur. It is how anyone can later tell what “broad support” meant when it did not.

Recognition asks for support without prescribing the route

The Recognition provisions make this distinction unusually clear. A candidate RIR cannot be recognised unless it demonstrates, to the satisfaction of the existing RIRs and ICANN, that resource holders in the proposed service region broadly support the candidate and will participate in its governance. The wider Numbering Community must also be committed to supporting it and participating in its policy process.

Those are substantive tests. The draft does not reduce them to a ballot total. Each existing RIR independently assesses the proposal. It may ask questions and conduct due diligence. It must send ICANN its recommendation, reasons, questions and received answers, as well as actual, potential or perceived interests connected to the outcome. ICANN publishes the proposal and assessments, subject to consultation about confidential redactions, and conducts its own review.

ICANN cannot recognise a candidate without support from at least two-thirds of the RIRs that submit assessments. A dissenting RIR whose service region or revenue would be materially affected may be excluded from that threshold, but only after written notice, ICANN’s reasoning and a fair opportunity to respond. Its input remains available on an informational basis.

This is not an evidence-free process. It creates records, attributed assessments and a high institutional threshold. Yet Article 2.7 means none of those actors is obliged by this document to open a consultation while determining whether broad regional support exists.

Support can be inferred from member resolutions, existing public records, letters, surveys, meetings, due diligence or a process created under another instrument. Some of those routes may be excellent. They are not interchangeable. A meeting room reveals intensity among attendees; a membership vote reveals a defined electorate; a survey reveals answers from respondents; operational records reveal behaviour. Each has a denominator and a selection effect.

If the resulting assessment merely says “broad support”, the reader cannot reconstruct which population, version or decision was tested.

Derecognition gives people a trigger, not a guaranteed hearing

The Derecognition route begins differently. Any RIR may submit a proposal. So may ICANN. A group comprising at least 25% of the subject RIR’s total members or 2,000 members, whichever is less, may also initiate one; each legal entity counts once even if it holds multiple votes elsewhere.

The proposal must identify the alleged non-compliance and provide supporting information. The subject RIR receives a reasonable period to answer. The other RIRs assess the case, ask questions, publish reasons and disclose interests. ICANN reviews the record and cannot derecognise without support from two-thirds of the RIRs, excluding the subject.

The draft also presumes rehabilitation first. Derecognition is not meant to be the first response to a fixable failure. That protection matters, especially when changing a registry’s status could move responsibility for services on which running networks rely.

But a member threshold authorises a proposal; it does not prove what the rest of the affected region thinks. The subject RIR’s response protects the institution’s procedural position; it is not the same as hearing resource holders, non-member operators, civil society, governments or technical communities. Published assessments show what decision-makers concluded; they do not necessarily show how affected communities could put evidence into the record.

Article 2.7 preserves the option to consult all of those groups. It also preserves the option not to. The difference should not disappear inside the final Notice of Decision.

Optional consultation has a serious defence

A universal consultation requirement can become a poor control. An urgent threat to registry integrity may not tolerate a long open-comment cycle. Litigation, personal data or security facts may lawfully limit disclosure. Five RIRs and ICANN operate under different corporate rules, and a single global procedure could duplicate a regional meeting or distort a process already required by local law.

Consultation also has familiar weaknesses. Participation volume is not representation. Organised groups can dominate an open list. A proposal written in technical and legal English can be formally open while remaining expensive to understand. Requiring “consultation” without specifying the evidence object can produce theatre: a window opens, comments accumulate, and no one can tell which changed the decision.

The draft’s alternative safeguards are real. Notices must be published. Assessments carry reasons. Questions and answers travel with the record. Interests are disclosed. The subject RIR can respond. Recognition and Derecognition both face institutional support thresholds, and Derecognition follows a rehabilitation preference.

These features explain why Article 2.7 need not be replaced with a referendum clause. They do not explain why the choice to consult should remain invisible.

Publish the choice, not a fiction of unanimity

A consultation-disposition receipt can be small. At each Recognition or Derecognition stage, it would identify the actor responsible for the assessment, the exact proposal version under review and the affected community. It would say whether a dedicated consultation was held, which other procedure supplied community evidence, or why consultation was not proportionate at that stage.

If consultation occurred, the receipt would record opening and closing dates, channels, eligibility or participation rules, the relevant denominator where one exists, and a disposition of material submissions. It need not expose personal data, privileged advice or security-sensitive facts. It needs to show how evidence crossed into the assessment.

If consultation did not occur, the receipt would give a bounded reason: urgency, duplication, legal restriction, adequate evidence already available or another stated ground. A later decision-maker could disagree, but it would know what choice had been made.

The receipt would not let attendees replace members, boards, RIR assessments or ICANN. It would not convert a stakeholder into a principal. It would prevent the reverse conversion: using the language of community support without identifying the community evidence behind it.

The recommended draft is not a live exercise of Recognition or Derecognition power. No source reviewed here establishes a pending proposal, a skipped consultation or an adopted rule. That uncertainty is part of the proper conclusion. Article 2.7 is a design choice visible now, before any case makes it politically expensive to examine.

Bottom-up governance does not require every decision to begin with an open microphone. It requires the route from affected people to authorised decision to remain inspectable. If consultation is optional, its use—or its omission—needs a named owner, an evidence trail and a reason.

Sources