Summary

  • Draft 5 retained a review-and-recovery structure already visible in Draft 4: contact, attempted correction, three months of public identification, recovery and a database update. Its real revisions formalised some complaint evidence, aligned the transfer route, rewrote appeal and reporting language, and changed authorship and acknowledgements.
  • Those revisions mattered procedurally. They gave a member more intelligible opportunities to answer, correct or transfer, and they acknowledged that appeals and reporting had to sit within ordinary law. They did not, however, prove that the RSA or another signed agreement authorised each review trigger and consequence.
  • AFRINIC’s defensible role was a narrow one: maintain an accurate uniqueness ledger, test concrete record and contractual discrepancies, and record valid corrections. It had no sovereign, police, penal, confiscatory or public-law adjudicative power, and policy language could not create any such power.
  • A sound private-registry model would bind every remedy to a specified resource and signed obligation, disclose evidence and reasons, provide usable cure and independent suspensive review, preserve historical records, accept proportionate responsibility for error, and protect portability and market alternatives.

L3 — The three-month cure was already on the page

The most revealing way to read Draft 5 is not as a clean break with what came before, but as a selective repair to an existing mechanism. Draft 4 had been submitted Under Discussion on 11 April 2017. It already contained the backbone that later gave Draft 5 its remedial appearance: AFRINIC would contact the organisation, try to correct an observed discrepancy or non-compliance, publish affected resources for three months if the issue remained unresolved, recover them after that period and update the prior holder’s database records. It also contemplated reallocation, an appeal and an annual report.

The three-month clock was therefore not an October invention. Neither were contact, correction, recovery or database updating.

That chronology matters because a casual account can grant Draft 5 more reforming force than its own revision history supports. The October text did not discover cure after an earlier proposal of immediate action. It retained a pathway that had already combined correction with a serious final consequence. The recorded changes were narrower: a new evidence-form paragraph after legal consultation; alignment of the transfer wording in section 13.4(B) with transfer policies; a rewritten appeal provision; a broader reporting formula; and changes to co-authors and acknowledgements.

These were meaningful edits, but they were edits to the machinery, not proof of the machinery’s legal foundation.

The intervention between the two drafts helps explain why those edits appeared. On 26 April 2017, AFRINIC staff and legal counsel assessed Draft 4 and identified questions that could not be answered merely by declaring a review desirable. Evidence might originate in several jurisdictions and would need reliability testing. The party under investigation needed a fair opportunity to be heard. Confidential sources created disclosure problems.

Arbitration raised questions of jurisdiction, procedure and reviewability; the assessment noted the possibility of Supreme Court annulment of awards as well as legal exposure, and it said an arbitration process would be required for implementation. Reporting raised its own concern about what applicable rules allowed AFRINIC to reveal. These were not objections to accurate records as such. They were warnings that a high-impact private process had to be institutionally and legally credible.

The AFRINIC-26 meeting record of 31 May 2017 adds another precise piece of context. It described Draft 4’s design in terms that included recovery of resources considered non-compliant and reallocation of recovered resources. It also recorded the co-chair decision that consensus had been reached and that the proposal would move to last call. That meeting action is relevant to the proposal’s progress. It is not the same thing as proof of adoption, contractual incorporation or operative authority. Draft 5’s own official status on 21 October remained Under Discussion.

Nothing in the material considered here establishes that Draft 5 was ratified, implemented, applied to a member or used to recover a resource.

Draft 5’s scope was broad in resource type but more selective in operational priority. It said the review could cover allocated or assigned IPv4 resources, IPv6 resources and autonomous system numbers. At the same time, it prioritised IPv4 and ASNs that could be mapped to two-octet ASNs. Coverage and priority should not be collapsed. The first described what fell within the proposed review universe; the second indicated which portions of that universe would receive earlier attention. The text did not say that IPv6 disappeared from scope, nor that every resource class would be pursued with equal urgency.

The same care is needed with the three review classes. A random review could draw from Medium and larger members, IPv6-only Large members and EU-AS members. A selected review could follow an internal report or a lack of contact. A reported review had different origins: the member itself could request one, or an evidence-backed complaint from the community could prompt investigation if AFRINIC staff judged the material sufficient to warrant it. A member asking AFRINIC to review its own position is not institutionally equivalent to an external complaint.

Nor is a complaint the same as an internal report, and lack of contact is not itself the same as evidence proving a contractual breach. Draft 5 placed these triggers within one policy design, but the different origins should carry different evidential and procedural demands.

Draft 5 also retained a 24-month limit that can easily be overstated. When a full reported review had been completed within the preceding 24 months, the same resource portfolio was excluded from another full review under that reported class. The shield applied to the same portfolio and to that class. It was not expressed as a universal immunity against random or selected review, a permanent repose period for the member, or a general limitation on every inquiry AFRINIC might undertake. Its narrower virtue was still real: repeated evidence-backed complaints could otherwise become a means of imposing perpetual review costs on one holder.

The provision offered some protection against that risk without answering whether the first review, or a review under a different class, was authorised.

The most visible response to the staff and legal concern about evidence was new paragraph 13.3.3(C). After assessing evidence submitted with a community complaint, AFRINIC staff could, at its sole discretion, require a sworn affidavit or a declaration before a Commissioner of Oath. Formal attestation can improve accountability. It may deter casual accusation, identify a person who stands behind a factual assertion and create a clearer object for later challenge. In a cross-border environment, however, form and truth remain different things. A sworn statement can be false, mistaken, incomplete or outside the declarant’s direct knowledge.

It does not resolve conflicts between jurisdictions, disclose confidential material to the affected member, establish a sound chain of custody or show that the asserted conduct breached a particular signed term.

Nor did the staff discretion itself become an independent evidential tribunal. Draft 5 allowed AFRINIC personnel first to assess whether complaint material warranted investigation and then to decide whether stronger attestation should be required. That may be practical triage for a private registry, but it combines gatekeeping and evidence-form decisions in the organisation that would also operate the review machinery. The paragraph is best understood as a procedural filter, not a certificate of truth.

A credible process would still need to separate the fact that a complaint had been formally made from the conclusion that its allegations had been proved to the standard necessary for an adverse record action.

The October text narrowed the stated grounds on which non-compliance could initiate the recovery path. It listed breach of AFRINIC policies and breach of the Registration Service Agreement, or other legal agreements between AFRINIC and the resource-holding organisation. That was more disciplined than broader formulations that could turn general information, incomplete contact or managerial suspicion into free-standing grounds for recovery. Yet the first category remained potentially expansive. “Breach of AFRINIC policies” could not be a blank cheque.

The label did not show which policy bound this member, when and how it entered the bilateral relationship, which conduct violated it, or what consequence the signed agreement allowed.

The distinction between a trigger and a proved ground is decisive. An internal report, lack of contact or an evidenced community complaint may justify asking a question. It does not by itself answer the question. Even a member-requested review may reveal an error without predetermining the contractual consequence of that error. The proposal’s control surface ran from selection and evidence assessment through finding, publication, transfer approval, recovery, database updating, possible reallocation and participation in the constitution of an arbitrator pool. Each stage increased the significance of the prior one.

The later the stage, the less acceptable it became to rely on an undifferentiated policy formula.

At the centre of the D5 mechanism was a two-step effort at correction. First, AFRINIC was to attempt contact and correct observed discrepancies or non-compliance towards the RSA. That is recognisably within a registry’s ordinary competence when it means verifying a contact record, establishing the status of an entry or asking a member to perform a concrete signed duty. Second, if the issue was not rectified and the resources had not been transferred to meet another AFRINIC-approved need under adopted policies, the proposal moved to public identification for three months.

During that interval, the organisation could seek compliance or transfer the resources to another member. D5’s revision record specifically characterised the transfer wording as an alignment with transfer policies.

This was a genuine attempt to create exits short of the final adverse act. Correction could restore an accurate record. A policy-compliant transfer could move operational control and registry recognition together rather than forcing an abrupt loss. A holder might use the three months to produce evidence, change a practice or reach another member. But each possibility depended on more than the existence of a date on a page. A cure is usable only if the member knows the exact allegation, sees the material needed to answer it, understands what action will count as correction and can complete that action in the time available.

A transfer is meaningful only if AFRINIC’s approval criteria are known and are not used to withhold the very exit offered as remediation.

After the three months, the text said the resource would be recovered and the former holder’s records updated in AFRINIC’s databases. It added that recovered resources could later be assigned or allocated under existing policies. The word “recovered” did substantial rhetorical work: it implied return to a proper administrative state. Yet the operational effect could be far greater than a tidy record correction. Registry recognition can affect a resource holder’s ability to demonstrate status, transact, plan routing support and assure customers. Reallocation would also introduce reliance by a later holder.

The database act therefore sat at the point where a private administrative sequence could produce asset-significant consequences.

Updating the former holder’s record should not be confused with erasing history. An accurate registry benefits from retaining provenance: who was recorded, what changed, when it changed, on whose decision and for what reason. Historical integrity makes error review possible and reduces the risk that a contested administrative conclusion is transformed into an apparently timeless fact. Draft 5 stated the database update, but the material does not establish a detailed historical-preservation rule.

A proper redline of the remedy should have distinguished the current status field from the archival trail and should have made clear that a disputed change did not disappear into an overwritten record.

The appeal rewrite was another material, incomplete response to the April concerns. Draft 5 said a dissatisfied reviewed member could appeal the result “within the four weeks”. It routed the appeal to arbitration under Mauritius’s Code of Civil Procedure. On request, AFRINIC could suggest a pool of knowledgeable community volunteers as potential arbitrators. The change moved away from treating an AFRINIC-defined internal pathway as sufficient and acknowledged that ordinary Mauritian civil procedure supplied the surrounding legal frame. That was an improvement in institutional vocabulary and possible accountability.

It did not turn AFRINIC into a court, and it did not establish that an arbitration arrangement had actually become operational.

The phrase “within the four weeks” was notably under-anchored. Four weeks from the first contact, the finding, the publication notice, the end of the cure period or some formal service of reasons would produce very different rights. The text, as recorded here, did not specify the starting event. It also did not establish that filing an appeal would automatically suspend public identification, recovery or the database update. Without suspension, a member might win a later argument yet suffer the operational consequences first. And AFRINIC’s ability to suggest knowledgeable volunteers on request was only that: a suggestion facility.

Knowledge of the community can aid technical understanding; a pool suggested by a party operating the challenged process is not proof of an independently constituted tribunal.

Finally, Draft 5 required an annual meaningful report on review activities in accordance with applicable laws and regulations. Broader and more lawful reporting language can improve oversight. It can reveal volumes, trigger classes, disposition rates, cure outcomes, transfer outcomes and appeal results without exposing confidential case material. The provision did not promise publication of member identities, source statements or every item of evidence, and it should not be read as though it did. “Meaningful” had to coexist with confidentiality, data protection and fair-hearing duties.

Aggregate institutional learning was the plausible aim; public trial by annual report was not.

Taken as a redline, Draft 5 therefore did something more modest and more defensible than either its champions or its critics might be tempted to say. It preserved a remedy sequence already in Draft 4 and repaired selected parts of the evidence, transfer, appeal and reporting architecture. It made the route look more like compliance remediation. It did not resolve the threshold warning embedded in the earlier assessment: a policy text could describe a process, yet only contract, consent and ordinary law could authorise a private registry to impose a consequence on a holder’s recognised resources.