Summary

  • AFPUB-2016-GEN-001-DRAFT01, published on 18 May 2016 by Amelina Arnaud, Ilunga Kabwika Serge, Jean-Baptiste Millongo and Wafa Dahmani with the status Under Discussion, proposed audits spanning all allocated or assigned resources and a sequence that could end in record removal, resource recovery and possible reallocation. It proves a proposed architecture, not an adopted rule or an action against any member.
  • The draft contained a legitimate administrative idea: AFRINIC could verify a specific registry fact, evaluate relevant evidence, notify a holder of a discrepancy and accept a correction. Its mistake was to connect that process to punishment-like consequences and an AFRINIC-defined final arbitration arrangement.
  • AFRINIC is a private technical bookkeeper and coordinator. Operational dependence on its ledger does not give it sovereignty, regulatory jurisdiction, police power, punitive power, confiscatory power or adjudicative power.
  • A proportionate system would use stable triggers, equal sampling, narrow evidence requests, secure handling, adequate cure time, reasoned findings, independent review and aggregate reporting. It would preserve the last verified operational state while a genuine dispute went to competent legal or contractual adjudication.

The proposal in its own time

The official title of the first draft was Internet Number Resources Audit by AFRINIC. The commissioning description calls it the first Resources Review draft, but that later vocabulary must not blur the document actually published. Draft 1 used the word “Audit”, carried the reference AFPUB-2016-GEN-001-DRAFT01 and was marked Under Discussion. Its four named authors were Amelina Arnaud, Ilunga Kabwika Serge, Jean-Baptiste Millongo and Wafa Dahmani. The revision note says that this first draft was posted to the RPD list. None of those facts establishes adoption, a vote, consensus, implementation or legal validity. They locate a proposition at a particular moment.

That distinction is essential because the draft was unusually ambitious in the sequence it put on paper. It proposed a way to select a member for audit, inquire into its use and records, classify an apparent failure, seek contact and correction, publish resources for recovery, remove the previous holder’s records, recover the resources, potentially reallocate them, and resolve a challenge through an arbitration process defined by AFRINIC. Read as a single administrative sequence, it can sound orderly. Read step by step, it crosses several institutional boundaries.

The starting concern was not imaginary. Internet number-resource records are useful only if they correspond closely enough to operational reality. Wrong contact information can obstruct urgent coordination. Contradictory records can make it harder to establish who is entitled to request a change. False or stale material can frustrate planning and disadvantage applicants where resources are finite. Draft 1 framed its problem in terms of finite resources, operational need, avoidance of stockpiling and section 4 of the Registration Service Agreement. Those were the proposal’s premises.

They were not proof that any member had stockpiled anything, that any particular record was false or that AFRINIC possessed an enforcement jurisdiction.

The institutional question is therefore not whether a registry should be blind. A registry which could never verify the entries it maintains would be a poor registry. The question is what follows from a verification and who may decide it. Checking whether a named contact still has authority is one act. Establishing what documentation supports an entry is another. Finding that two records cannot both be correct is another. Publishing an accusation, extinguishing an operational position, assigning the resource elsewhere and declaring the dispute finally decided are different acts again.

Draft 1 placed too many of them in the hands of the same private institution.

What the first draft would have put under review

The proposed scope extended to all allocated or assigned resources, while giving priority to IPv4 and to ASNs mappable to two-octet ASNs. It did not state an audit population, an audit cadence or a sample size. Instead, it described three routes into review: random, selected and reported audits.

For random audits, the draft named the classes Medium and above, IPv6-only Large, and EU-AS. Those labels should be kept exactly as the text gave them. Supplying a modern definition or reconstructing a classification system would add facts the first draft did not establish. Even without such definitions, random selection has an intelligible benign purpose. If the eligible strata and selection method are declared in advance, sampling can test record quality while reducing the opportunity to single out a disliked member. But the value of randomness depends on design. A label alone does not reveal the size of a stratum, the odds of selection, exclusions, repetition limits or whether the draw can be reproduced.

Selected audits could begin after an internal report or a lack of contact. Both may point to a real registry problem. Staff may encounter conflicting evidence in ordinary work, and prolonged inability to reach the person authorised to maintain a record may leave the registry unable to verify it. Yet each trigger also needs boundaries. Draft 1 did not identify the threshold for an internal report, the person who would decide whether it justified an audit, the number or form of contact attempts, or any recusal rule. “Lack of contact” can describe an evidence gap, but it is not by itself proof of wrongdoing.

An unreachable contact might reflect an obsolete address, staff turnover, a failed delivery channel or some other mundane discontinuity. The proper first response is to verify and repair the contact chain.

Reported audits could follow a member’s own request or a community complaint said to require investigation. Self-request is the least troubling route: an organisation may have good reason to seek formal verification before a merger, operational change or internal clean-up, although the draft did not provide such examples. A complaint is more difficult because an allegation can be useful evidence, strategic pressure or simple error. Draft 1 published no complaint evidence threshold, confidentiality rule, standing rule, burden of proof, standard of proof, reviewer-independence rule or recusal rule.

It therefore described an entry point without supplying the procedure required to distinguish a credible signal from an attempt to weaponise administrative attention.

The draft also offered a non-exhaustive list of possible non-compliance. It included an unjustified absence from the global routing table; breach of policy; breach of the RSA or another legal agreement with AFRINIC; evidence that an organisation no longer operated while its blocks had not been transferred; and unauthorised transfers under policy. These were proposed examples, not findings against a named organisation. They did not establish a universal measure of proper resource use, and they did not collapse every contractual, technical and corporate question into the same evidential test.

Routing visibility shows why precision matters. A prefix absent from the global routing table is an observation about what a particular view of routing discloses at a particular time. It is not automatically a conclusion about entitlement or misuse. A network may have operational reasons for not announcing a resource globally; visibility may also vary with routing configuration, observation point or timing. Draft 1 qualified the example with unjustified, which at least acknowledged that absence and misconduct were not synonyms. But it did not publish a test for justification. Without a stated question, relevant evidence and a fair opportunity to explain, an apparently objective technical signal can be made to carry a conclusion it cannot support.

The same discipline is needed for the other examples. A suspected policy breach, a suspected contractual breach, evidence that an organisation no longer operates and a suspected unauthorised transfer raise different questions. Each may involve a different document, effective date, decision-maker, evidential burden and appropriate forum. An audit can compare registry information with credible evidence and record a discrepancy. It cannot make all disputes clerical simply by placing them in an audit section.