Summary

  • After AFRINIC-3, AFRINIC placed four proposals into one 15-day mailing-list last call before Board approval. The governing transitional process expressly permitted final changes and amendments, so the meeting did not freeze the texts. Yet it did not say who would judge a new material objection, whether substantial edits required renewed consensus, or what showing could stop forwarding.
  • The visible exchange changed the procedural record, not the four policy texts. Alan Barrett challenged the opening notice’s claim that a vote had taken place; Alan Levin recalled that one might have occurred but said he could be wrong; AFRINIC chief executive Adiel Akplogan then stated that there had been no vote and that the process rested on consensus.
  • No substantive new policy-text objection appears in the archived sequence before Ernest Byaruhanga announced expiry and forwarding. That silence makes the hard counterfactual untestable. It proves neither that objections were powerless nor that any objection would automatically have reopened the process.
  • The last call was valuable as an asynchronous evidentiary safeguard, but it conferred no public mandate. AFRINIC remained a private bookkeeper and coordinator; its meeting, list, staff and Board were internal process actors, not a legislature, sovereign authority or owner of the resources its records helped coordinate.

The decisive event did not occur in the Cairo meeting room. It occurred after the chairs had been stacked away, when the description of what happened inside that room was exposed to people who had not been there. On 9 February 2006, Ernest Byaruhanga, AFRINIC’s Registration Service manager, opened a bundled 15-day last call for four number-resource proposals that the AFRINIC-3 report had described as receiving meeting consensus. The proposals were due to go to AFRINIC’s Board for review after that interval. The first public challenge did not attack any of their substantive provisions.

It attacked the notice’s assertion that the meeting had voted.

That correction captures both the worth and the limit of last call. A physical meeting had produced an apparent conclusion. An asynchronous list gave the conclusion an outside checkpoint, lowered some of the costs of travel and presence, and created a written trail. It worked: the public description of the meeting changed. But the underlying rule did not explain what would happen if an outsider raised a material objection to the policy itself. The four proposals touched direct or end-user address access, temporary resources, eligibility for autonomous system numbers, and the upstream allocation of IPv6 space from IANA to regional registries.

Operators and resource users therefore had practical reasons to care whether a late intervention could amend, delay or stop a proposal. The archive answers the procedural-description question. It leaves the material-objection question open. Reading further matters because institutions are often judged by their most visible step, while the real allocation of control lies in what that step obliges anyone to do.

The safeguard between meeting consensus and Board action

The transitional policy-development text in force for this batch placed last call after meeting endorsement and before Board ratification. It said that anyone could propose policy, that an open policy-working-group list would discuss proposals for at least 30 days before a public meeting, and that meeting endorsement would rest on consensus rather than majority vote. Once a proposal had meeting consensus, the process announced a 15-day last call in which the community could suggest final changes and amendments. Only after that stage did the Board ratify and adopt policy for AFRINIC’s use.

For the narrow question here, two features matter more than the rest of the sequence. First, the public meeting was not formally the last opportunity to affect the text. The explicit invitation to suggest final changes and amendments meant the proposal remained open after the room had expressed consensus. Someone unable to travel to Cairo still had a defined interval in which to put an objection, correction or edit into the record. Second, the instrument described an opportunity but not a complete decision rule. It did not identify who accepted or rejected a proposed amendment.

It did not state whether a material amendment had to return to the list for another discussion period, go back to a meeting, or receive a fresh consensus assessment. It did not say whether one well-founded objection could defeat the prior finding of consensus. Nor did it name an actor required to publish a reasoned closure determination.

Those omissions should be treated precisely. They are absences from the governing six-step text available in the record, not proof that AFRINIC staff had no internal practice or private discussion. Equally, an undocumented practice cannot be supplied by assumption. The public process told a participant that final changes were welcome. It did not tell that participant the consequence of making one. That gap matters because access to a comment channel and power to alter the path are not the same institutional property.

AFRINIC itself described this early structure as bottom-up and participative in its recognition-application materials. Those materials also said its first public policy meeting had about 120 registered participants, including representatives of more than 30 African local internet registries. Such statements document how AFRINIC presented its process and the participation it reported. They do not establish that the participants were an electorate for every resource holder, operator, state or user in Africa. They also do not change the legal and institutional nature of the organisation.

AFRINIC was a private, membership-based number-registry coordinator administering records and a policy process. Openness can improve that process. It cannot convert attendance into public authority.

What the Cairo record established—and what it did not

The AFRINIC-3 meeting report places the event at the Mena House Oberoi in Cairo on 13 and 14 December 2005. Byaruhanga’s later opening notice gives the dates as 12 and 13 December. The discrepancy cannot responsibly be smoothed away; both are institutional records, and the available evidence does not provide a basis for silently choosing one. What can be said is that the meeting occurred in Cairo in mid-December and that its policy discussions were led by Alan Barrett, identified as a policy-working-group member, and Byaruhanga, identified as Registration Service manager.

The report described four proposals later included in the February batch as having received meeting consensus. They concerned end-user or direct assignments, temporary assignments, changes to ASN-assignment criteria, and the allocation of IPv6 address space from IANA to regional registries. Each was to return to the policy list for 15 days before final Board approval. These surfaces could affect how networks acquired resources, met eligibility conditions or planned around upstream allocation. They therefore carried operational and commercial significance without needing to be treated here as four separate policy-merits debates.

Meeting consensus was an input to the next stage, not a referendum and not a transfer of ownership. The AFRINIC-3 report does not provide a denominator for the policy session, a denominator for membership, a ledger showing who spoke, or a map showing which participant had authority from which principal. An open door is useful; it is not a theory of representation. The people in the room could contribute technical knowledge, evidence, objections and practical judgment. Their participation could help AFRINIC decide what rules to use in its own coordinating work. It could not bind all absent operators by sovereign command.

The physical setting nevertheless created a concrete governance problem. Attendance carries costs: travel, visas, permission from an employer, language, time and familiarity with procedure can all determine who is present when apparent consensus is formed. A meeting-only finalisation rule would embed those costs into the ability to influence resource policy. Last call reduced that risk by separating the provisional meeting conclusion from Board action and exposing it to a wider, asynchronous audience. It was a safeguard against room capture precisely because it denied the room exclusive control of the final evidentiary record.

One notice, four proposals and an unstable description

At 11:34:47 SAST on 9 February 2006, Byaruhanga opened one last call covering all four proposals. He addressed the wider community, summarised issues discussed at the meeting and said the proposals would be sent to the Board for review after 15 days. On temporary assignments, the notice recorded consensus as subject to suggested edits. That detail reinforces the point that the post-meeting text was not conceptually frozen.

The notice also described the meeting outcome using the language of both consent and a vote. That formulation triggered the only visible exchange during the call. On 11 February, Barrett replied that there had been no vote, although consensus appeared to exist. On 13 February, Alan Levin said he thought a vote might have occurred before discussion, but expressly conceded that he might be wrong and recalled only two voting members being present. Later that morning, AFRINIC chief executive Adiel Akplogan stated that there had been no vote, that policy was not a voting matter under the process, and that adoption rested on consensus.

Akplogan’s message settled AFRINIC’s stated procedural characterisation in the archived exchange. It did not settle a question of public-law legitimacy, because none was available for the list to confer. Levin’s uncertain memory should not be promoted into an established vote count: Barrett and Akplogan said no vote occurred, and Levin qualified his own recollection. Yet his intervention remains important evidence of how quickly an oral meeting can produce a fragile retrospective record. Three people who had reason to know the process did not begin with an identical description of what had happened.

The list enabled that disagreement to surface and the organisation’s formal position to be made explicit.

This was not a cosmetic success. Calling an outcome a vote implies a different aggregation mechanism from describing it as consensus. A vote has a count, an electorate and a threshold; consensus depends on discussion, objection and judgment. If the process required the latter, an official opening notice invoking the former could mislead readers about how the proposals advanced. Without last call, the Cairo room’s apparent consensus might have gone directly to Board review while the incorrect account remained in the public record.

The asynchronous checkpoint therefore produced observable information value even though it did not alter the four substantive texts.

The exchange also exposes an evidentiary asymmetry. AFRINIC’s meeting report and emails prove what AFRINIC and named participants recorded or said. The institution’s description of “community”, “consensus” or participation cannot prove a mandate that the process itself did not establish. Official status makes a record strong evidence of an institutional act; it does not make every institutional interpretation a constitutional fact. The correction should be credited because it improved the record, not inflated into evidence that the corrected process represented a continent.

What archived silence can and cannot bear

The visible archive under this subject contains four messages during the call: the opening notice and three replies about vote versus consensus. The next archived message is Byaruhanga’s closure announcement. No substantive new objection to the text of any of the four proposals appears in that sequence. This yields a firm but narrow finding: in the observable batch, the last call corrected the process description but did not visibly change the policy language in response to a new policy-text objection.

It does not yield a reliable answer to the harder research question. Because no material objection arrived in the observed sequence, the archive never tested the missing decision rule. One cannot infer that AFRINIC would have ignored such an objection. One also cannot infer that any objection, however weak, would have stopped forwarding. The proposal might have been edited, recirculated, returned to a meeting, abandoned or forwarded with an explanation. The available evidence does not choose among those possibilities.

Silence has practical meaning in a coordination process, but only after the process defines what has been asked, from whom, against which text and by what deadline. Here the open channel and invitation to amend were real. Yet the absence of a version hash, issue ledger and published adjudication rule limits what silence can demonstrate. It may support an administrative inference that no visible participant raised a substantive concern. It cannot prove universal acceptance by people who did not participate, did not see the notice, could not engage in time or never authorised anyone in the forum to speak for them.

The strongest case for AFRINIC’s approach deserves full credit. In 2005, an early registry combining at least 30 days of open list discussion, a public meeting, a final 15-day list interval and Board review had a practical, proportionate sequence. A process cannot remain permanently open because it has not pre-written an answer to every hypothetical. When a public opportunity is offered and no substantive objection arrives, administrators need to be able to close the stage and continue. The list reduced travel barriers, created an archive and, in this case, improved the record. Those are substantial virtues.

The answer is not to deny those virtues. It is to separate administrative sufficiency from political authority and to observe that this particular safeguard was never stress-tested. Open participation plus silence can reasonably support an internal decision to proceed when the organisation has explained the rule. It does not create a sovereign mandate, and it does not eliminate the value of specifying what a future material objection would trigger. Indeed, the correction of the vote claim makes the case for explicit closure stronger: the archive demonstrated that descriptions coming out of the room could be wrong.

Expiry announced, reasons unrecorded

The end of the interval is itself recorded inconsistently. The later policy history for temporary assignments gives 24 February 2006 as the end of last call. Byaruhanga’s subsequent closure message says the call expired at 12:00 SAST on 25 February, while misprinting the year as 2005. The evidence does not justify normalising those records into one exact deadline. For a participant deciding whether a submission was timely, the difference between the two dates would be material. For historical analysis, the discrepancy is another sign that a time-bounded safeguard needs an unambiguous timestamp published at the start.

On 7 March 2006 at 12:58:11 SAST, Byaruhanga announced that the last call had expired and that the policy working group would forward the four proposals to the Board for approval. He is therefore the identifiable opener and closure announcer in the public record. The AFRINIC-3 report had also identified him as one of the people leading the meeting discussion. Those facts establish an observable administrative role. They do not establish that the transitional instrument designated him or his office as the final judge of post-call consensus.

The closure notice did not include an objection ledger, a redline of text changes, a reasoned assessment that consensus still held, a named chair’s determination, or an explanation of how a material new objection would have affected forwarding. That statement concerns the content of the public closure record; it does not claim that no internal deliberation occurred. The distinction is important. Governance should be evaluated through what affected parties can verify, not through speculation about what staff may have considered privately.

Board approval then stood as a separate internal corporate step. The later AFRINIC-4 report said the Board had approved the policies concerning direct or provider-independent and critical-infrastructure assignments, temporary address assignments, changed ASN criteria, and IPv6 allocation from IANA to regional registries. The available record does not contain the underlying resolution, exact ratification date, vote or approved redline. The later report supports the fact of reported approval, and no more.

Board ratification gave AFRINIC an internal adoption event for the rules it would use in coordinating number-resource records. It did not turn the Board into a legislature, the policy working group into an electorate or the list into a source of public law. Nor did it confer police, prosecutorial, confiscatory or punitive power. AFRINIC did not own Africa’s number resources by keeping the registry or administering the process. It coordinated uniqueness, records and allocation practices within a private institutional arrangement. That narrower description is not an insult to the function.

It is the condition for assessing the function accurately.

A room is evidence, not a people

The correct authority map starts with the operators and other principals who bear the consequences of network-resource decisions. They face routing, customer, contractual, capital and continuity effects. AFRINIC’s role is to keep records and coordinate a process intended to make resource use unique and interoperable. It does not become the network operator merely because its records matter to operations. The meeting, mailing list, staff and Board are components of that coordinating machinery. Their decisions can matter greatly without possessing sovereign force.

Heng Lu’s central institutional distinction applies without dilution: a room is not a people, a mailing list is not a legislature, consensus is not ownership and Board ratification is not public law. Participation is valuable because it supplies evidence, expertise, warnings, objections and technical discipline. It should be welcomed for those properties. It cannot be transmuted into authorisation from absent principals merely by being called open, bottom-up or multistakeholder.

That distinction also clarifies the present roles of organisations outside the event. NRS had no operational role in AFRINIC’s 2005-2006 meeting, list, closure or ratification sequence. Its legitimate domain is to advocate, research, convene and represent members who have explicitly authorised it. That member authorisation is different from claiming to embody an undefined community. It does not make NRS an operator of the registry, RPKI, WHOIS or RDAP, appeals, elections, settlement, custody or continuity, and it does not give NRS power to close AFRINIC policy.

LARUS likewise had no process role in this event. Its relevance is the operator and business-continuity lens: the organisations running networks and carrying commercial consequences are not replaced by people assembled in a meeting room. Being an operator or continuity provider does not itself confer policy authority. BTW.Media also had no role in running the meeting or process. Its place is the reality and intelligence layer—making institutional acts, records, contradictions and consequences legible—while the AFRINIC documents and list archive remain the evidence for what occurred.

Reporting does not substitute for those primary records, and neither reporting nor operational exposure creates public power.

These layers prevent a common category error. Representation must be traceable to a principal; operations must be judged through actual continuity consequences; public understanding must be built from inspectable records; and registry administration must remain a coordinating function. None of the layers may absorb the others. A registry cannot claim the operator’s losses as its own mandate. An advocate cannot claim to operate the registry. An operator cannot convert exposure into legislative power. A publication cannot turn analysis into an institutional act.

The narrow verdict

The 15-day last call gave people outside the AFRINIC-3 room a real opportunity to place comments and proposed amendments into the official policy discussion before Board review. The transitional text kept that channel formally open, and the archived exchange demonstrates that it could change something important: it corrected the account of how the meeting had reached its conclusion. In the counterfactual with no last call, that error might have survived into Board review without a final asynchronous challenge.

But “opportunity to speak” cannot be equated with “proved ability to alter the outcome”. The process did not explain how a new material objection would affect the text or forwarding decision, and the observed batch supplied no such objection with which to test the mechanism. The closure appeared publicly as a staff announcement of expiry and onward transmission, not as a reasoned, independently reviewable determination. The deadline itself is inconsistent across the later records, and the subsequent Board approval is reported without its underlying resolution or redline.

The fair conclusion is neither cynicism nor reverence. The safeguard was real, useful and worth preserving. It widened the evidence base beyond one physical room, reduced some participation costs and corrected a consequential procedural misdescription. It was also incomplete. Its under-specified closure rule left the most important contingency—what happens when a late objection is material—unresolved. Improving that rule would strengthen private coordination. It would not create sovereign legitimacy, because no mailing-list design can turn attendance into mandate over those who never authorised the forum.