Summary

  • The recommended RIR Governance Document would let a group representing at least 15% of an RIR’s total Members, or 1,000 Members if that is lower, request a Compliance Review. It counts each legal entity once, regardless of extra voting weight elsewhere.
  • LACNIC’s public election numbers distinguish authorised voters, participating organisations and votes cast. None is automatically the “total Members” denominator in the draft, so the implementation procedure needs a dated, auditable member count before a request arrives.

A thousand votes sounds like a number. A thousand members is an institutional definition.

That difference sits inside section 4.2(b)(i)(C) of the ASO AC’s recommended third draft of the RIR Governance Document. The clause would allow a group of an identified registry’s Members to request a Compliance Review when the group reaches at least 15 per cent of the RIR’s total Members or 1,000 Members, whichever is lower. It immediately adds a rule that is easy to skim past: each legal entity counts as one Member for this threshold, even when the entity carries multiple votes under the registry’s rules for other matters.

This is not yet an operative right. The ASO’s 1 September notice calls version three a recommended draft and says further opportunities for input, and any additional updates, will be announced or coordinated by the NRO Executive Council. Nor does the number alone compel a review. Under the draft text, a requester must identify the particular obligations in section 4.1 said to be at issue and demonstrate sufficient materiality. A member-led request must also show that at least one member made adequate efforts through the RIR’s available processes. ICANN then assesses whether initiation is warranted.

The arithmetic nevertheless matters because it is the first gate. A request cannot be evaluated on its merits until somebody can establish who was entitled to join it and how many qualifying members existed on the relevant date. “Total Members” could be mishandled as a count of resource accounts, voting contacts, eligible voters, ballots, voting units or corporate groups. The legal-entity sentence rules out some shortcuts, but it does not by itself produce a clean list.

LACNIC offers a useful illustration without supplying the answer. Its 2025 Board election result reports 12,501 authorised voters, 1,207 organisations that voted and 3,127 votes cast, with participation labelled at 10 per cent. Those are three plainly different quantities. LACNIC’s membership page explains why: the number of votes assigned to a member depends on the amount of address space managed by the organisation. Its bylaws chapter also distinguishes membership categories and their rights.

None of those election figures should be reverse-engineered into the draft’s denominator. An authorised voter may be a voting entitlement rather than a unique legal person; an organisation that voted is a turnout measure; votes cast reflect weighted voting. The public result is valuable precisely because it makes the units visible. It shows why “1,000” cannot be copied from the nearest election dashboard.

An implementation procedure therefore needs what might be called a denominator certificate. It should state the record date; the membership categories included; whether suspension, arrears or pending admission changes eligibility; how direct and national-registry relationships are treated; and how the same legal entity is recognised across multiple accounts or contacts. It should publish the total after deduplication and identify the officer or independent party responsible for the calculation. Privacy does not require publishing the entire member roll.

A signed aggregate, rules, versioned methodology and a channel for private verification can make the threshold auditable without exposing every member.

The requesting side needs a matching receipt. Each signatory should be linked to a qualifying legal entity as of the record date, with duplicates and withdrawn support resolved under published rules. The bundle should separately name the section 4.1 provisions, the evidence of materiality and the earlier attempt to use LACNIC’s own processes. Those are different tests. Combining them into one opaque acceptance decision would make it impossible to know whether a request failed on numbers, procedure or substance.

The rationale report records the effort to make the compliance framework clearer. Clarity at this boundary is unusually valuable. A member trigger is credible only if supporters know what counts before they organise, the registry cannot revise the denominator after seeing the names, and ICANN can explain its disposition in terms that both sides can reproduce.

There is no evidence here that LACNIC faces a Compliance Review, that any group is collecting signatures or that the registry is out of compliance. The current question is more prosaic and more timely: whether the machinery will be measurable on the day someone first tries to use it.

The next NRO EC update should therefore be read for six items: a record date, an eligible-Member definition, a legal-entity deduplication method, a secure evidence channel, a public receipt for the resulting count and an ICANN notice stating why a request did or did not cross each gate. If those pieces appear together, 1,000 becomes a threshold. Without them, it remains typography.