Summary

  • Exactly three local policy proposals were late at AFRINIC-2: temporary assignments, end-user IPv4 assignments, and an ASN-eligibility amendment. The IXP, root-server and ccTLD provisions belonged inside the end-user proposal, while the IPv6 IANA-to-RIR item followed a separate global-policy sequence.
  • Public archive dates left only 11 elapsed calendar days for the temporary proposal and 10 for the other two before the 27 April policy session, against a stated minimum of 30 days. AFRINIC-2 discussed the texts but did not move them into meeting consensus, the 15-day last call or Board adoption; official records place consensus at AFRINIC-3 and the downstream states later.
  • The result was procedural restraint, not rejection on the merits. The checked record names no person who issued the disposition, preserves no vote or motion, proves no motive beyond insufficient notice, and quantifies no operator harm.
  • A notice floor can reduce surprise and improve the evidentiary record without creating political sovereignty. AFRINIC remained a private, non-governmental registry and ledger coordinator whose process could govern its own allocation criteria, not a legislature with public-law punishment power.

A young registry meets an old problem

On 27 April 2005, in Maputo, the decisive fact was not how loudly anyone spoke. It was how little time had passed. AFRINIC's own account of its policy development process said that a proposal had to spend at least thirty days on the open policy mailing list before it proceeded to the face-to-face meeting for endorsement through consensus. Consensus, in the institution's description, meant general agreement rather than a majority vote. If consensus was reached, a separate fifteen-day last call followed. Board ratification and adoption came after that.

The three proposals before AFRINIC-2 had not reached the first threshold. Alain P. Aina's temporary-assignments text appeared in the public archive on 16 April 2005 at 15:05:21 UTC. Mark Tinka's ASN-eligibility amendment appeared on 17 April at 16:37:08 UTC, followed twenty-seven seconds later by his end-user IPv4 proposal at 16:37:35 UTC. Measured by calendar dates to the policy session on 27 April, the first interval was eleven days and the other two were ten. The exact session time is not published, so a finer elapsed-hour calculation would be false precision.

It could only shorten the completed interval, not turn ten or eleven days into thirty.

That shortfall is large enough to survive ordinary counting disputes. The temporary proposal was at least nineteen days short under an elapsed-day calculation; the two 17 April posts were twenty days short. If the measurement ended at the meeting's opening on 26 April rather than the policy day on 27 April, the temporary proposal would have accumulated only ten calendar days. Inclusive rather than exclusive counting cannot plausibly bridge the gap. The public record may leave details uncertain, but eligibility under the stated floor is not one of them: zero of the three proposals had completed thirty days.

AFRINIC nevertheless discussed them. This distinction is the centre of the event. Notice did not operate as a gag rule. Participants could question the texts, expose operational concerns and influence later revisions. What the meeting did not do was treat that conversation as the consensus state that unlocked last call and Board adoption. The procedure separated the right to be heard from eligibility for institutional effect.

That separation can sound formalistic until the proposals are read as operating rules rather than abstractions. They concerned temporary number resources, provider-independent IPv4 assignments and ASN eligibility. Their terms could influence routing, applicant evidence, fees, membership status and registry work. A discussion held ten days after public posting could be useful and still be too early to carry the consequence assigned to meeting consensus. AFRINIC-2's defensible act was to harvest information without pretending that the promised notice interval had elapsed.

The count is three

The first reconstruction problem is deceptively simple: what was actually deferred? The answer is exactly three proposal objects.

The first was Alain P. Aina's proposal for temporary assignments and allocations. Its initial scope covered temporary unicast IPv4 and IPv6 addresses and AS numbers for documented, time-bounded activities. The submitted text proposed a one-month lease, renewable on application, alongside return and administrative-fee questions. Meeting discussion, as AFRINIC later summarized it, raised practical concerns: whether one month was too short, whether temporary address space might acquire a poor routing reputation, whether a reserved block was appropriate, and how fees should work. Those were material implementation questions.

They were comments on a proposal, not evidence that its one-month term or any immediate-withdrawal language had become policy.

The second was Mark Tinka's proposal for IPv4 assignments to end users in the AFRINIC service region. It addressed provider-independent or end-user eligibility, a proposed /24 minimum, utilisation tests and the handling of additional assignments. It also included a critical-infrastructure section covering Internet exchange points and core DNS operators, including root-server and ccTLD uses. The proposal's internal structure matters because the AFRINIC-2 report visually formatted “IXPs, Root Server or ccTLD Operators” in a way that can resemble a separate heading in a list of policies.

The archived submission resolves the ambiguity: those terms were a section of the end-user IPv4 document. They do not form a fourth late proposal.

The third was Tinka's amendment to the ASN eligibility criteria. It proposed adding AFRINIC membership in good standing to existing criteria built around a unique routing policy and multihoming. AFRINIC's meeting report also summarized a restriction on local internet registries assigning AS numbers to other organisations. The report says this item received little discussion. “Little” is not “none,” and neither is it consensus. The proper state is that the amendment was among the three discussed proposal objects, but no AFRINIC-2 consensus or adoption is recorded.

A different item must remain outside this count. The IPv6 allocation-from-IANA-to-RIRs presentation appeared separately in the meeting report as an NRO or global-policy matter, before the section dealing with the late local proposals. Alan Barrett's located submission of the global proposal into AFRINIC's archived policy process is dated 1 August 2005, months after AFRINIC-2. The complete circulation history before August is not established by the sealed record, so that date should not be made to prove more than it can.

It is sufficient, together with the report's structure, to keep the IPv6 global-policy sequence distinct from the three April local submissions.

These two counting corrections prevent opposite errors. Splitting the end-user document's critical-infrastructure section inflates three into four. Folding the global IPv6 presentation into the late local set mixes a different authority chain and chronology into the event. Both errors would turn document formatting into institutional fact. Proposal identity must follow the actual submissions and their process path.

Five states, not one meeting

The March 2005 recognition application that AFRINIC submitted to ICANN is evidence of how AFRINIC described its then-current process. It set out a sequence. Anyone could propose a policy. The open working-group list supplied discussion and comments. After at least thirty days, a proposal could go to an open face-to-face meeting for endorsement through consensus. A fifteen-day last call followed meeting consensus. The Board then ratified and adopted the policy for use.

This source proves what AFRINIC represented its process to be. It does not independently prove that the process was neutral, universally representative or politically legitimate simply because AFRINIC used those terms in an institutional submission. Official documents are strongest when used to establish an institution's own words, reported acts and declared procedures. They are not independent witnesses to the truth of self-legitimating claims.

Within that boundary, the process map is clear. Public-list notice, meeting discussion, meeting consensus, last call and Board adoption were separate states. Implementation was another practical state after institutional adoption. Conflating any pair changes the event.

Public-list notice meant that a text was available in the designated forum for the required interval. It did not measure support. A busy thread would not prove broad agreement, and a quiet thread would not erase the calendar. Nor does archive presence identify every affected network that actually received notice. It records availability and whatever messages were preserved, not a complete population of principals.

Meeting discussion allowed people in the room to test a proposal. The AFRINIC-2 report records substantive questions about the temporary and end-user texts and little discussion of the ASN change. That interaction mattered because it could shape a later version. It did not cure the missing pre-meeting interval. The very purpose of advance notice was to let people outside the room inspect the text before a consensus determination, not merely to comment after the meeting had announced a result.

Meeting consensus was the next institutional state, contingent on timing eligibility. AFRINIC described it as general agreement rather than a majority vote. The checked AFRINIC-2 record does not preserve a vote, a motion or a consensus call for these proposals. More fundamentally, the report's treatment places them outside that state: it repeats the thirty-day requirement and says that the end-user policy would be discussed and approved or rejected at the next member meeting. Later records locate consensus for the three objects at AFRINIC-3, not in Maputo.

Last call was downstream of consensus and lasted fifteen days under the stated process. It was not an interchangeable second bucket of notice. A post-meeting interval occurs after the room has expressed general agreement; a pre-meeting interval occurs before that agenda-setting act. Using the later period to cure missing prior notice would reverse who receives time before an institutional signal has been sent. The official policy histories place the eventual last call in February 2006, after the later meeting consensus.

Board ratification and adoption were later again. The checked record contains no AFRINIC-2 Board adoption of the three proposals. It also does not expose a final Board resolution and exact ratification date for each later text. Official histories label the policies implemented, but that label does not fill the missing resolution trail. Nor may ambiguous May 2005 “Date” fields on legacy policy pages safely be read as implementation dates, because those dates precede the separately recorded December consensus and February last call.

This is why saying merely that AFRINIC-2 “deferred” the proposals can be both convenient and dangerous. It describes the outcome, but it may imply a discrete ruling whose author and words are known. The record supports a reconstruction: the report applied or observed the notice floor, allowed discussion and withheld adoptable-consensus treatment. It does not identify the person who made that determination, the formal role in which it was made, or the exact language used in the room.

What the meeting record does—and does not—say

AFRINIC's meeting report says recent policies and amendments had been proposed about two weeks before the meeting. The exact archive dates yield only eleven and ten calendar days to the 27 April session. That modest difference may reflect conversational rounding, an unrecorded circulation route or something else. No explanation is established. It does not affect the core conclusion because both “about two weeks” and the exact public-post calculation fall far below thirty days.

There is also a one-day discrepancy for the temporary proposal. Its archived message header says 16 April; a later AFRINIC history describes it as first posted on 17 April. A timezone convention or later archival migration might explain the mismatch, but the checked sources do not say. The correct response is not to choose a theory. The header supplies the high-confidence public timestamp used for the eleven-day calculation, while the later institutional history's different date remains disclosed. Either date leaves a decisive shortfall.

The meeting report records discussion across the three proposal objects. For temporary assignments, questions about duration, routing, reserved space and fees went directly to operational design. For end-user IPv4 assignments, the record describes routing-table and interim-request concerns. For the ASN amendment, it notes little discussion. These summaries establish the subjects AFRINIC reported as raised. They are not a verbatim transcript, a complete speaker list or evidence of how many participants held each view.

The report says more than 100 people attended the two-day AFRINIC-2 meeting. That figure cannot be repurposed as the denominator for the policy session, still less as an electorate of affected African operators. It does not say that every attendee was present for each discussion, authorised by a network to speak, or representative of all potential applicants and resource holders. Attendance is evidence of presence. It is not a transfer of mandate.

No located source names a chair, moderator, staff member or Board actor who issued a procedural ruling. No exact motion, objection, vote or consensus-call transcript has been located. The official report may reflect a decision taken in the room, a procedural understanding, or later report synthesis; the present record does not allow those possibilities to be separated. Accountability would be stronger if the institutional actor and reasoning were named. Their absence does not erase the observable state: the proposals were discussed, did not enter consensus at AFRINIC-2, and appeared in later consensus records.

The motive boundary is just as important. Insufficient notice is the recorded procedural reason. There is no evidence that AFRINIC or any participant intended to suppress, favour, embarrass, protect or discriminate against a proposer or constituency. The later progression of the proposals would make a simple merits-rejection story especially misleading, but it still does not prove that every actor was neutral. Motive is unknown. A serious reconstruction distinguishes an absent proof of bad faith from a positive proof of good faith; the record offers neither.

The impact boundary also remains narrow. No identified holder is shown losing an allocation, route, ASN, RPKI object, reverse-DNS delegation, contract right or payment because the proposals waited. There is no quantified backlog, denied request, extra fee, outage, routing failure or investment delay. The texts plainly touched matters with economic and technical consequences, but a mechanism is not an observed harm. The history supports procedural delay and operational stakes, not a damage estimate.

The price and value of waiting

The strongest case against AFRINIC-2's restraint is practical. AFRINIC was a young registry facing real requests. The proposals were concrete rather than ceremonial. More than 100 people attended the broader meeting; those present did discuss the issues; and the ideas later advanced. Temporary event operators might need short-lived resources. End users might need a local path to provider-independent IPv4 space. ASN applicants might need clear eligibility criteria. Why should a formal clock prevent useful action when the room could reach provisional consensus and use a later last call to collect objections?

That case identifies a real cost. The calendar rule traded speed for reviewability. A local AFRINIC rule for end-user requests did not become available in April, and the meeting report records concern about requests pending the next meeting. Temporary-use applicants did not receive the proposed formal local route at AFRINIC-2. The interval from 27 April to the opening of AFRINIC-3 on 13 December was 230 elapsed days. That is a meaningful duration for an operator, even though the record does not show who waited or with what result.

But the efficiency argument cannot change the sequence AFRINIC had published. The thirty days came before meeting consensus. The fifteen-day last call came after it. A post-consensus opportunity to object does not fully substitute for a pre-consensus opportunity to assess the text, because an announced consensus changes expectations, agenda power and the burden on dissent. No emergency exception allowing provisional consensus was located. Treating the clock as optional at the first difficult moment would have taught participants that the written notice floor could be compressed whenever the meeting considered a proposal useful.

Nor did AFRINIC-2 face a binary choice between immediate adoption and operational paralysis. Its report attributes to CEO Adiel A. Akplogan the interim position that the relevant former RIR's policy would continue to apply to end-user requests until AFRINIC adopted its own. That statement is evidence of an operating bridge, not adoption of Tinka's late proposal. The legal and operational basis for choosing among predecessor policies is a separate subject and is not resolved here. For the narrow event, it shows only that discussion without adoption did not necessarily leave staff with no rule at all.

The meeting also gained information by discussing the proposals. The temporary text's one-month term could be challenged as too short. Routing reputation and reserved-block design could be examined. Fees could be questioned. The end-user proposal's /24 minimum and critical-infrastructure conditions could be tested for routing-table and eligibility effects. This is a better use of a late agenda item than either silence or premature effect: conversation could improve later drafts while the institution preserved the distinction between feedback and decision.

The later chronology supports that state separation without proving that every day of delay was necessary. AFRINIC-3's report lists the three proposal objects among items receiving meeting consensus and says they would return to the list for fifteen days before Board approval. Later histories record a fifteen-day last call from 9 to 24 February 2006 and mark the policies implemented. The complete April-to-December version diffs are not in the checked record. Therefore, later consensus cannot be projected backwards onto the April wording.

The eventual destination confirms that AFRINIC-2 was not a merits rejection; it does not establish that the proposal texts remained unchanged.

The AFRINIC-3 report itself requires bounded use. It contains an internal inconsistency concerning an unrelated four-byte ASN item, which it places in one passage among consensus items and elsewhere describes as not discussed toward consensus because of insufficient notice. That contradiction lowers confidence in using the report for every item in its universe. It does not overturn the three-proposal chronology, which is consistently reflected in their individual official histories. It is a reminder that an official record can be authoritative evidence of institutional reporting without being infallible.

Notice as discipline, not mandate

The thirty-day floor deserves credit for what it did. It constrained the institution at a moment when speed was attractive. It created a readily auditable calendar condition. It gave operators and other affected parties more time, at least in principle, to inspect terms concerning routeability, prefix size, utilisation, membership and fees. It kept the meeting from using its own discussion as proof that the wider notice promise had been satisfied.

That is process discipline. It is not sovereignty.

AFRINIC was and remains, for the purpose of this authority analysis, a non-governmental, not-for-profit registry coordinator and membership company administering number-resource records and services. Its policy process could organize how it adopted criteria for resources and registry services within lawful, corporate and contractual limits. The ability to maintain a unique ledger and process requests did not make it a state, a legislature, a regulator of all African networks or an owner of the internet's number resources.

The distinction matters most when proposal language sounds coercive. A temporary-assignment draft might contain return or withdrawal conditions. An ASN proposal might connect eligibility to membership in good standing. Those provisions could become contractual or registry criteria after a valid process. At AFRINIC-2, they were only proposed clauses. They were not sanctions that had been adopted, and nothing in the record establishes a sovereign power to punish. Registry coordination can support accurate, unique records; it cannot manufacture public-law authority by calling a meeting a community.

Likewise, openness does not automatically establish representation. Anyone's ability to propose is an access rule, not evidence that every affected principal authorised the list or the room. Thirty days of public availability is better than ten. It may reduce surprise and allow objections to surface. But time on a mailing list does not demonstrate that an IXP, an end-user network, an event organiser or an ASN applicant saw the text, could participate effectively, or consented to every resulting rule. Notice is necessary discipline in the institution's chosen process; it is not a plebiscite.

This doctrine does not make the notice rule meaningless. The opposite follows. Once sovereign rhetoric is removed, procedural commitments become more important, not less. A private coordinator's legitimacy depends substantially on bounded function, predictable procedure, accurate records and fidelity to the authority it actually possesses. It cannot rely on the mystique of “community” to cure shortcuts. Obeying the thirty-day minimum was a modest, verifiable act of institutional self-restraint.

The proposals also sat largely in the part of registry policy with the strongest functional case: criteria for future assignments from pools the registry administered. A coordinator needs working rules for evaluating requests. Yet even future-allocation criteria can impose costs and shape routing behaviour, so the boundary between technical coordination and broader control must be watched. If a clause were to threaten existing resources or operate as punishment, the authority and safeguards required would be more demanding.

Nothing about the AFRINIC-2 discussion resolves that larger issue or prejudges the technical merits of a /24 minimum, a one-month term or multihoming criteria.

What can be concluded at the evidence cutoff

The event admits a firm conclusion precisely because the conclusion is narrow. AFRINIC had published a thirty-day prerequisite for meeting consensus treatment. The public archive dates show only eleven days for the temporary proposal and ten days for the ASN and end-user proposals by 27 April. The meeting discussed all three. It did not record consensus, last call or Board adoption for them. The three objects later appeared in AFRINIC-3's consensus set and then in later last-call and implemented histories.

Everything beyond that needs qualification. The named procedural decision-maker is unknown. The exact words used to describe the disposition are unknown. The full policy-session participant set and speaker record are unknown. Any private circulation before the public posts is unknown and, in any event, would not by itself satisfy a rule described as public-list discussion. The complete version chain from April to December is unknown. The final Board resolutions and adoption dates are not present in the checked public record. The meaning of the legacy May 2005 display dates is unresolved. Operator costs and outcomes are unquantified.

The source condition also deserves candour. At the evidence cutoff, direct checks of three legacy AFRINIC policy-detail slugs returned 404, although indexed official archive copies preserved the relevant history and text. The core April finding does not depend on those later pages: the contemporaneous meeting report, the ICANN-hosted recognition application and the mailing-list archive establish the notice rule, timestamps, proposal identities, discussion and non-adoption state. The legacy histories are used only for bounded later chronology and should be rechecked before a future reseal.

Several discoveries could change parts of this account. An April or May 2005 Board resolution adopting an exact draft would require a fresh evidence seal and reconciliation with both the stated prerequisite and the later consensus record. A transcript could name the person who made the procedural call or show that the report compressed a more complex exchange. A complete version chain could reveal how much the April discussion changed the texts. Request records could demonstrate tangible costs of waiting. None of those possibilities licenses invention now.

The historical lesson is therefore neither “procedure defeated policy” nor “community consensus worked.” It is more useful. A registry separated discussion from effect when its own notice clock had plainly not run. That restraint preserved room for later review and did not prevent the proposals from progressing. It also left applicants under inherited rules for longer, with consequences the available record cannot measure. The act was institutionally defensible because it honoured a stated condition, not because AFRINIC possessed sovereign authority.

Eleven days was not thirty. Ten days was not thirty. Three proposals were discussed, and none crossed the next process boundary in Maputo. Keeping those plain facts separate from motive, mandate and mythology is the difference between a reliable institutional history and a story assembled from suggestive headings.