Summary

  • AFRINIC-32 supplied important context for the Resource Transfer Policy project, but it took place before Draft 4 was submitted on 5 October 2020 and therefore could not establish assent to that later text.
  • Rough consensus properly gives co-chairs room to weigh reasons rather than count hands, yet that discretion requires a version-specific ledger showing each material objection, its evidence, the clause it reached, the response and the reason for its disposition.
  • Participation figures should provide context by independent entity and role, not operate as a voting threshold; neither repetition, silence nor meeting atmosphere can demonstrate the breadth of affected support.
  • Better documentation would have reduced compatibility, diligence, delay and appeal costs without granting a lone persistent speaker a veto over progress.
  • Whatever the quality of the process, AFRINIC remains a private technical registry and coordinator. Its consensus can organise a voluntary service rule, but it cannot legislate for absent holders, operators, customers or states.

L3 — The fourth text and the missing denominator

The decision object changed on 5 October

The practical difficulty began with chronology. The Resource Transfer Policy project had entered AFRINIC’s policy process on 30 October 2019. By the time of AFRINIC-32 on 16 September 2020, participants had a history of proposals, arguments and revisions to discuss. That history mattered. It showed that inter-regional transfer rules were not a merely stylistic exercise. They raised questions about needs assessment, legacy resources, the respective roles of source and receiving registries, outgoing transfers and whether counterpart registries could operate with the proposed terms.

Yet chronology also establishes a boundary. Draft 4 was submitted on 5 October, after AFRINIC-32. The September meeting could reveal which concerns had been prominent, how participants understood the policy project and where proponents thought compromise might be possible. It could not reveal whether the meeting accepted the exact October wording. A room cannot consent prospectively to unknown text. Nor can a chair safely treat approval of a policy direction as approval of every later mechanism used to express it.

That distinction may seem formal until one asks what a consensus call actually advances. It does not advance a general aspiration to make transfers possible. It advances a text capable of governing how a private registry service will handle particular requests. A phrase about the role of the source registry, the receiving registry, need or legacy status may alter who must show what, which conditions must be satisfied, or whether another registry regards the procedure as compatible with its own. When material wording changes, the decision object changes with it.

Draft 4 therefore needed its own evidentiary life. Earlier support could be carried forward only where the record showed that the relevant proposition remained the same. Earlier objections could be treated as answered only where the new wording, evidence or explanation demonstrably reached them. General approval of inter-regional coordination was not a blank cheque for the fourth text. General opposition to transfer was not automatically a material objection to every clause. The co-chairs’ task was to distinguish the two with enough precision that another reader could follow the judgement.

The official archive identifies AFPUB-2019-V4-003-DRAFT04, records 5 October 2020 as its submission date and records its movement towards the consensus and last-call stage. The September minutes provide the preceding discussion context. Public reporting in the available record also says ARIN and APNIC had regarded wording in Draft 3 as incompatible in particular applications. Those facts do not decide whether Draft 4 solved the problem. They show why a version-specific answer was necessary. Compatibility feedback addressed to one draft cannot simply be assumed to bless the next.

A quieter discussion is not necessarily convergence

Long policy discussions produce several kinds of silence. A participant may have been persuaded. Another may believe a revised clause resolves the concern. A third may have stopped responding because the exchange is repetitive. A fourth may lack the time or confidence to continue. A fifth may consider the forum incapable of protecting the operational interest at stake. Others may never have joined at all. These states sound identical on a mailing list: no new message appears.

The same ambiguity affects a meeting. An energetic exchange can create a vivid impression of engagement while representing a small group of recurring participants. A show of general comfort may indicate broad convergence, or merely that the people still in the room have heard the argument before. Repetition can enlarge message volume without enlarging support. Fatigue can reduce message volume without resolving the reason for dissent. Neither pattern is evidence of bad faith by itself. Both are reasons to refuse easy inferences from atmosphere.

For Draft 4, the relevant question was not whether discussion had gone on for a long time. It was whether the reasons bearing on the exact fourth text had been dealt with. The difference is crucial. Duration is a property of the process; disposition is a property of the argument. A proposal can be old while containing a new and unsettled clause. It can be noisy while having a narrow participation base. It can become quiet while leaving a technically important concern intact. It can also remain noisy after the substantive concern has been fully answered because one speaker refuses to accept the answer.

Rough consensus exists partly to deal with this last possibility. It is not a requirement that every person declare satisfaction. Co-chairs must be able to say that an objection has been understood, tested and answered even when its author continues to repeat it. Otherwise persistence becomes a veto. But the legitimacy of that classification depends on showing the path from objection to answer. Without the path, observers see only a conclusion: consensus was found. They cannot tell whether the conclusion reflects convergence, attrition or a sensible refusal to reward repetition.

The 2021 appeal matters in this limited way. AFRINIC’s appeal register records a formal challenge to the Draft 4 consensus determination, and the appellant disputed substantive changes, revision disclosure and the handling of objections. The filing is an advocate’s account, not a neutral finding. It does not establish that every criticism was correct, that every objection remained unanswered or that the co-chairs acted improperly. It does establish that the decision record remained contestable enough for a participant to invoke a formal challenge.

That later contest makes the quality of the earlier evidentiary bridge important; it does not tell us which side ultimately deserved to prevail.

Counting the right things without turning them into votes

There is no reliable assessment of breadth unless the unit being described is clear. Ten messages might come from ten independent operators, or from one person restating the same concern. Five meeting interventions might represent five distinct operational roles, or five people from the same organisation. A person may speak as a resource holder, an employee, a consultant, a technical expert or an interested individual. Those roles can all contribute useful knowledge, but they do not have the same relationship to the consequences of the proposed rule.

This is the denominator problem. It is not solved by announcing a majority threshold. Rough consensus need not become fifty per cent plus one, and the meeting need not pretend to be a legislature. The denominator is contextual information: how many independent entities and affected roles appear in the record, how concentrated participation is, and whether apparent agreement spans different operational positions. It tells readers how much confidence to place in a claim of breadth. It does not mechanically command the outcome.

The distinction between a stakeholder and a principal helps. A stakeholder may be affected, informed or interested. A principal has authority to decide for a defined interest. A meeting attendee, mailing-list contributor, database contact, member of staff or consultant can offer evidence. Appearance in an open process does not authorise that person to bind absent holders, operators, customers, end users or governments. Describing participation honestly therefore requires more modest language than describing a regional mandate.

For Draft 4, an adequate participation account would have separated speakers and messages from independent entities, and entities from roles. It could have identified, without turning personal prominence into power, whether concerns came from operators facing transfer procedures, resource holders exposed to continuity or compatibility risk, technical participants examining the wording, or advocates advancing a general policy position. It should also have marked repeated interventions so that frequency did not masquerade as diversity.

Such an account would not prove that the larger group was correct. A technically decisive objection may come from one well-placed operator. A large number of expressions of support may rest on a misunderstanding. Qualitative judgement remains necessary. Yet the reverse is also true: chairs cannot claim broad convergence without disclosing enough context to show what “broad” describes. Reasons decide the treatment of an objection; participation context disciplines claims about reach.

The record available for this analysis does not contain a participant-by-independent-entity denominator for the Draft 4 determination. It also does not include the complete last-call correspondence or the full co-chair reasoning statement. That is a limit on what can responsibly be concluded here. It would be wrong to infer that no fuller records ever existed. It would be equally wrong to invent counts, reconstruct support from fragments or declare that all objections either survived or disappeared.

What can be said is narrower and more useful. The decision advanced one exact fourth draft. The meeting that supplied much of the policy context occurred before that draft. Compatibility concerns had been reported around Draft 3. A later appeal disputed the handling of changes and objections. In those circumstances, a transparent denominator and a clause-level disposition record were not decorative extras. They were the means by which a qualitative judgement could be distinguished from inherited momentum.

The first test for readers

Anyone assessing a similar consensus call should begin with a paired document: the frozen text and its complete redline against the version most recently discussed. Both should remain accessible together. The next question is whether each material objection has an identifier and reaches a stated clause. A third question is whether any external compatibility view identifies the exact version examined. A fourth is whether participation has been described by independent entity and role, not merely by messages, hands or impressions.

These checks do not demand unanimity. They demand that the object of judgement remain stable long enough to be judged. If the text changes after the meeting, the changed text needs a new comparison and a renewed opportunity for focused response. If the chair concludes that an objection has been answered, the record should show the answer. If silence follows, the most that can safely be said is that no new recorded objection appeared during the relevant interval. Silence should not be converted into an affirmative declaration of consent.

This is the practical line between convergence and fatigue. Convergence can be explained clause by clause: the text changed, evidence answered the concern, the concern was withdrawn, narrowed or reasonably classified as non-substantial. Fatigue is inferred from dwindling participation without such an explanation. The two may coexist; a discussion can converge on most questions while wearing people down on one. A careful assessment does not require mind-reading. It requires an inspectable map from the fourth text to the reasons for advancing it.