Summary

  • Draft 2 retained the proposal’s random, selected and reported opening lanes, renamed an “Audit” as a “Review”, and changed the reported-lane test from a complaint that “requires investigation” to one that “warrants investigation”.
  • The changed verb invited a preliminary judgement, but the draft did not publish the evidence, criteria, decision-maker, reasons, notice, scope or conflict rules by which that judgement would be made.
  • That omission mattered before any finding of non-compliance: opening a review could consume staff and counsel time, expose operational information and give the registry leverage through uncertainty about possible downstream consequences.
  • A legitimate record-integrity function needed a narrower gate: verifiable allegations, objective relevance and materiality criteria, separation of intake from decision, proportionate scope, written reasons and a reviewable opening record.

L3 — One changed verb and an opening rule left unwritten

Draft 1 said that an audit could be initiated in the reported lane when a member asked to be reviewed or when the community supplied a complaint that “requires investigation”. Draft 2 used the same basic lane, but referred to a complaint that “warrants investigation”. Around that one verb, the architecture remained strikingly familiar. There were still three ways in: random, selected and reported. The proposal’s title moved from “Internet Number Resources Audit by AFRINIC” to “Internet Number Resources Review by AFRINIC”.

Yet the text did not explain what made a complaint warrant investigation, who made that judgement, or what record had to support it. The immediate practical consequence of crossing the line was not merely semantic. A member could have to divert people, collect records, expose information and plan around uncertain registry action before anyone had established non-compliance.

That is the narrow change at the centre of Draft 2. Draft 1, published on 30 May 2016, called the process an audit. Its section 3.3.3 listed random, selected and reported initiation. Draft 2’s recorded revision history lists a change of policy name, the addition of acknowledgements and a rephrasing of section 3.3.3. Comparing the reported clauses supplies the precise alteration: “requires” became “warrants”. The random categories remained materially the same. So did the selected lane, which could rest on an internal report or lack of contact. Draft 2 did not replace a loose collection of triggers with a detailed opening code.

It retained the three-lane arrangement and recast one threshold in language that sounded evaluative without saying how the evaluation worked.

Even the date requires care. AFRINIC’s policy list and the revision history on the draft record 5 August 2016. The header on the Draft 2 detail page displays 9 August 2016. The four-day discrepancy cannot responsibly be resolved by choosing whichever date makes a cleaner narrative. It is more accurate to say that the institutional record places Draft 2 in early August, with 5 August supported by the index and revision history and 9 August displayed on the detail-page header. The discrepancy does not change the wording comparison, but keeping it visible is part of the discipline this subject demands.

A study of an unwritten decision rule should not begin by smoothing away an inconsistency in the documentary record.

The title change is similarly incapable of carrying more meaning than the text gives it. “Review” may sound less accusatory or punitive than “Audit”. It may better describe a registry checking the accuracy of records and a member’s private contractual compliance. But a softer noun does not answer the questions that govern entry. Who receives a community complaint? Does that person merely log it, or also decide that it warrants investigation? Must the complaint identify a particular registry fact, allocation-policy discrepancy or contractual obligation? Must it attach evidence capable of verification?

How serious must the alleged discrepancy be? None of those questions is answered by replacing the name on the door.

The verb does suggest a possible intention. “Warrants” can naturally be read as requiring a preliminary judgement rather than treating every complaint as self-executing. In the strongest benign reading, the edit was an effort to say that receipt alone was insufficient: someone would need to decide that investigation was justified. That reading should be taken seriously. AFRINIC had legitimate technical reasons to discover stale registry entries, unreachable members, organisations that had ceased operating, policy discrepancies and unauthorised transfers.

A system confined to problems volunteered by the affected member could overlook hidden errors. Random checks and credible reports can therefore serve record integrity.

But an intention to require judgement is not the same thing as a rule for exercising judgement. “Warrants” locates a decision point while leaving its contents blank. It does not say whether the threshold is evidential, prudential, operational or reputational. It does not specify whether a plausible allegation is enough, whether corroboration is required, or whether materiality matters. It does not distinguish a complaint identifying an objectively checkable registry inconsistency from one expressing general dissatisfaction with a member.

The word can carry the idea of justification, but it does not publish the standard by which justification is assessed.

Nor did Draft 2 identify the institutional actor. The sealed record shows no distinct intake function and no separate opening decision-maker. That omission matters because receiving a complaint and deciding to impose a review are different acts. Intake is administrative: preserve the submission, confirm what it alleges, screen for completeness and identify any obvious conflict. Opening is consequential: decide that the allegation and its evidence justify compulsory attention from the member. When the text does not separate those functions, the same person or team may be understood to perform both.

That may be administratively convenient, but convenience is not a substitute for accountable design.

AFRINIC’s embedded staff assessment saw part of this problem. It accepted the use of complaints in principle, yet said it was unclear who would determine whether a complaint warranted investigation. It interpreted the decision as a matter for AFRINIC staff discretion and warned about unjustified complaints and the workload for staff and members. Those observations are important because they are contemporaneous institutional evidence that the operative words did not settle the question. They are not an independent judicial finding, and staff’s interpretation cannot manufacture authority that the proposal failed to define.

They nevertheless confirm that ambiguity at the gate was not a merely theoretical objection invented after the event.

The staff assessment also raised other practical concerns within the wider scheme. It questioned global route visibility as a policy requirement, identified tension between deleting records of prior holders and existing practice, and noted privacy problems with public reporting. It estimated three to six months to create an appeal system, followed by two months for process. Those concerns show that implementation would have demanded more than changing terminology. They also help explain why the opening decision cannot be isolated from institutional capacity.

If investigation, reporting and appeal consume scarce time, a vague gate does not only burden the subject of a weak complaint. It also transfers the cost of screening bad or strategic submissions to the registry and its wider membership.

Still, the analysis must not outrun the evidence. Draft 2 was under discussion. The record does not establish adoption or implementation. It does not show that AFRINIC opened a review under the draft, targeted a member, or used the ambiguity selectively. There is no basis here to infer a malicious complainant, bad faith by staff or an actual disclosure burden borne by an identified operator. The institutional risk follows from the design of the proposed gate, not from an accusation about conduct. That distinction is essential: a rule can be inadequate even when every person expected to operate it intends to act fairly.

The difference between proposed design and proven application is especially important because the wider Draft 2 text contemplated serious downstream movement. It described regular reviews of compliance with the Registration Service Agreement and allocation policies, prioritised some kinds of resource holding, and provided a non-exhaustive set of possible non-compliance indicators. After contact and an opportunity to correct a discrepancy, unresolved matters could move towards recovery.

The draft contemplated a three-month publication period before recovery and reallocation, an appeal involving volunteer arbitrators defined by AFRINIC, and annual reporting that would name reviewed members and state their compliance status.

Those features are relevant only to show why opening carried weight. They are not an invitation to turn this inquiry into a study of recovery, revocation or appeal design. A gate matters in proportion to the process it enters. When the far end of a proposed review could affect resource records, public reporting and contractual relations, the first decision attracts leverage even if those consequences remain conditional and distant. Yet it would be equally wrong to speak as though opening mechanically produced recovery. Draft 2 contemplated steps in between, including contact and an opportunity to correct discrepancies.

The analytical point is narrower: potential downstream consequences make uncertainty at the threshold economically significant.

The official contractual evidence has to be handled with the same restraint. A later AFRINIC redline, dated 31 October 2017, says that it shows changes from the January 2016 Registration Service Agreement. It preserves baseline language in Article 4(b) concerning information reasonably requested by AFRINIC, assistance and co-operation during investigation or review, and possible private contractual consequences if a member failed to comply. It also preserves relevant Article 4(c)(iii) language. This helps identify the contractual environment in which the proposal was written.

It does not establish the exact signed agreement of every member, and it does not convert a private information-and-co-operation clause into sovereign investigative power.

That boundary is not rhetorical. AFRINIC is a private, membership-based technical bookkeeper and coordinator. Its registry function is useful because Internet number records require reliable administration and co-operation across operators. Its influence over records can be substantial. But usefulness and influence do not make it a government, regulator, police force, prosecutor, confiscator or court. A policy proposal cannot confer public-law authority, and a private membership contract cannot do so either.

The legitimate question is whether AFRINIC could conduct a narrow record and contractual check within private relations, not whether it could exercise a general power to investigate whomever the community complained about.

The distinction clarifies why trigger design is more than procedural decoration. An objective trigger connects the registry’s action to a specific technical or contractual discrepancy. A vague trigger connects action to institutional discretion. The former keeps the body near its bookkeeping task; the latter risks allowing the language and burden of investigation to expand beyond that task. The concern is not the word “investigation” by itself. It is the absence of a published link between the signal received, the technical role, the evidence required and the decision made.

AFRINIC described its policy-development process as open, transparent and participatory, with publication of drafts, mailing-list discussion and recorded version histories. That process provides useful visibility and a route for comment. It is part of the evidence that allows the narrow D1-to-D2 comparison to be made. But participation does not fill a substantive blank. People can openly discuss a text whose decision rule remains vague. A version history can faithfully record that a clause was rephrased without explaining how staff should apply the new verb.

Procedural openness cannot create sovereignty, and it cannot substitute for objective criteria at the point where a member is made to respond.

Later developments must also remain outside the frame. Safeguards introduced in later drafts cannot be backdated into Draft 2. The evidence here does not support attributing an evidence-backed complaint rule, a 24-month safe harbour, a formal opening notice or a cure ladder to the August text. Nor can later disputes, litigation or governance conflict be used as hindsight proof of what Draft 2 meant. The clean comparison is less dramatic and more revealing: three lanes stayed; the process was renamed; the reported-lane verb changed; the rule for applying it was not written.

That reconstruction produces a precise institutional diagnosis. Draft 2 did not necessarily lower the threshold, and it cannot safely be said to have raised it. “Warrants” may sound more selective than “requires”, but natural-language impressions do not establish a legal or operational delta when neither term is accompanied by criteria. What changed was the phrasing through which discretion would be expressed. What remained undefined was the evidentiary and institutional machinery needed to discipline that discretion. The proposal pointed at a judgement but did not show the reader how the judgement became justified.