Summary

  • On 27 April 2005, amid discussion of a proposed IPv4 end-user policy, the AFRINIC CEO advised that end-user provider-independent requests would be handled under the policy of the member’s respective former RIR until AFRINIC adopted its own policy. The surviving record supports that narrow interim instruction; it does not support blanket inheritance of every predecessor policy for every request class.
  • The instruction answered a genuine continuity problem, but it was not fresh local adoption. The proposal had been posted only about two weeks before the meeting, while AFRINIC’s transitional process required at least 30 days of open list discussion, face-to-face consensus, a 15-day last call and Board ratification.
  • “Respective former RIR” was a historical registry-relationship selector, not an explained merits test. The record does not resolve applicants with no predecessor, more than one predecessor, corporate succession, disputed histories or resources administered through another registry, and it identifies neither a version crosswalk nor a reason-and-review procedure.
  • AFRINIC is a private ledger operator, bookkeeper and technical coordinator. It has no sovereign, legislative, police, punitive, confiscatory or transnational regulatory authority. A defensible borrowed rule could preserve uniqueness, accurate records and operational continuity; it could not create ownership of number resources or a self-renewing mandate.

A live queue met an unfinished rule

The institutional problem on 27 April 2005 was small enough to fit inside a meeting question and important enough to reveal the character of registry power. AFRINIC-2 had opened the previous day at the Telecomunicações de Moçambique conference and training centre in Maputo. More than 100 participants were reported. Earlier that month, an IANA report had recorded AFRINIC’s recognition and identified ARIN, APNIC and the RIPE NCC as the incumbent regional registries that had served different parts of Africa. The meeting report also recorded a formal handoff from those three organisations on 27 April.

Those facts explain why a predecessor rule could be available to borrow. They do not confer sovereignty, title, public-law jurisdiction or a legislative mandate.

The immediate issue was the proposed IPv4 policy for direct assignments to end users. Its substantive tests were already concrete. A general minimum assignment of a /24 was contemplated. An applicant could show efficient use of at least a /25 already, or an immediate need for at least half of the requested space. A later request for an additional assignment would be tested against utilization of 25 per cent immediately and 50 per cent within one year. These were not merely abstract aspirations. They were criteria that could shape the evidence an applicant assembled, the questions staff asked and whether a request appeared eligible.

Yet the proposal had arrived about two weeks before AFRINIC-2. The meeting record therefore presented a timing mismatch. The organisation had begun operating as the regional ledger, applicants could have end-user provider-independent requests requiring attention, and the proposed local rule had not had enough time to complete the process that AFRINIC itself had published. Participants asked what would happen between AFRINIC-2 and the next member meeting, at which the policy would be discussed and approved or rejected by members.

The CEO’s recorded answer was that AFRINIC would enforce the policy from the respective former RIR of the member until the AFRINIC policy was adopted. In its own institutional vocabulary, “enforce” described how staff would process a request against a rule. It did not turn AFRINIC into a police power. Nor did the answer transform a newly discussed proposal into adopted local policy. The most accurate description is recorded CEO advice or an interim operating instruction for one request class.

That classification matters. The minutes do not identify a member vote on the instruction, a consensus finding adopting it, a Board resolution authorising it, or an amendment to the policy-development process. They record an operational answer to an operational question. Calling it “policy” without qualification would erase the distinction between keeping a service running and completing the procedure announced for making a local rule. Calling it “law” would be more misleading still.

The scope must be kept just as narrow. The question concerned end-user PI requests: applications by end users for provider-independent IPv4 space. The answer does not prove that AFRINIC imported every allocation, assignment, membership or administrative rule from ARIN, APNIC and the RIPE NCC. It does not establish a blanket inheritance code for every African request. The predecessor handoff is context for why the selector existed, not the subject of this analysis; the mechanics of moving records, staff responsibilities, services or number pools belong elsewhere.

So do the separate question of unused-resource recovery, Board-election history and a general institutional chronicle. IPv6 advocacy is also outside the frame. The issue here is the rule used in a specific administrative gap.

Continuity was not local adoption

The gap becomes clear when the instruction is placed beside AFRINIC’s transitional policy-development process. Dated 30 June 2004 and later marked obsolete, that process allowed anyone to submit a proposal. It then required at least 30 days of open discussion on a mailing list. The proposal would proceed to face-to-face consideration, where consensus meant general agreement rather than a simple majority vote. A 15-day last call followed. Board ratification and adoption came after those public stages.

The end-user proposal had been available for only about two weeks before AFRINIC-2. Even without reconstructing the exact posting hour, it had not satisfied the minimum 30-day list period by the time of the Maputo discussion. The later stages had not all occurred either. The fresh local rule was therefore unfinished when the CEO gave the interim answer.

This is more than procedural arithmetic. The four stages performed different functions. List discussion gave people time to see the text, test its assumptions and prepare objections. Face-to-face consensus offered a point at which general agreement could be assessed. Last call gave the wider list a final opportunity to catch defects or disagreement after the meeting. Board ratification supplied the corporate act by which the private association adopted the service rule. Compressing those stages into the fact that a room had discussed a proposal would collapse notice, deliberation, revision and adoption into attendance.

At the same time, none of those steps is sovereign lawmaking. A policy meeting does not represent every operator, resource holder, customer, state or other affected principal merely because participation is open. Consensus among those present is a governance method inside a private coordination body, not a source of public authority over absent parties. Board ratification is an act of corporate governance, not legislation. The value of the process lies in notice, intelligibility, opportunity to object and documented adoption within the service relationship. Its value does not depend on pretending that AFRINIC is a state.

Borrowing a predecessor rule therefore occupied an intermediate category. It was neither random discretion nor completed local adoption. It was a temporary compatibility measure: continue processing a particular kind of request under an earlier known ruleset while the replacement moved through the local process. That category can be legitimate, but only if its limits are kept visible. Continuity is the purpose. The predecessor document is the temporary input. Local adoption is the endpoint. None of the three should be allowed to masquerade as the others.

The distinction also prevents recognition language from doing work it cannot do. IANA’s 8 April report proves that a recognition act was recorded and that the three predecessor registries were identified. The handoff language in the AFRINIC-2 record proves that registry functions and records were described as passing to AFRINIC. Official terms such as “authority” or “custody” remain institutional vocabulary for a coordination transition. They do not prove sovereignty over persons, ownership of addresses or public-law jurisdiction across a continent. A chain of recognition among technical institutions can help coordinate a unique ledger.

It cannot bootstrap a transnational legislature.

The selector was historical, not geographical

The key phrase in the CEO’s answer was “respective former RIR of the member.” That is a relationship selector. It points backward to the registry that had previously served the member. The surviving report does not tell staff to choose by the applicant’s country, place of incorporation, customer base, routing origin, prefix provenance, physical network, preferred ruleset or any other merits criterion. Replacing the recorded selector with one of those alternatives would be invention.

For a member with one clear and continuous predecessor relationship, the phrase had an intuitive administrative meaning. Staff could identify the previous registry, retrieve its rule, and preserve the treatment on which the member had recently relied. The apparently simple case supplies much of the instruction’s practical appeal. It makes the transition feel less like a new gate and more like a continuation of a known service relationship.

But the selector becomes unstable at its edges. What rule applied to a new AFRINIC member with no previous relationship to any of the three? What happened when an organisation had dealt with more than one predecessor—perhaps through different operations—before consolidation? How should staff treat a renamed company, a merger, a split, or another form of corporate succession? Which record controlled if the historical relationship was disputed? What if the request concerned resources that had been administered in another registry’s system even though the member’s main relationship lay elsewhere?

The meeting report supplies no resolution for any of those cases.

Those are not exotic questions added to manufacture doubt. They are direct tests of the selector’s stated form. A rule based on historical affiliation requires a method for proving affiliation. A rule addressed to “the member” requires a method for connecting the current legal or operational identity to the former registry record. A rule that borrows another institution’s text requires a decision about which text and which date count. If those methods are not written down, staff discretion moves from applying a rule to choosing the rule that will be applied.

The clean minimum would have been a short selection protocol. First, determine whether the applicant was already a member with one unambiguous predecessor relationship. Second, identify the exact predecessor document and version in force for that member immediately before the handoff. Third, state that only the relevant end-user PI eligibility and evidence provisions were carried across, unless another provision was specifically named. Fourth, tell the applicant which selector and version were used and why. Fifth, isolate ambiguous-history disputes for review instead of silently choosing the most restrictive or most convenient corpus.

Sixth, explain how comparable applicants could understand differences caused by their distinct predecessor histories.

The minutes do not show such a protocol. They publish no crosswalk of applicable documents, no version dates, no treatment of conflicting provisions, no common evidence standard, no reason-giving rule and no appeal path. They do not state the instruction’s exact implementation start time. They do not explain the precise mechanics of its sunset. These omissions do not prove that staff acted unfairly, and they do not prove that no operational guidance existed outside the surviving record. They define what the public record cannot establish.

Three corpora, not one inherited custom

The predecessor policies were not interchangeable expressions of a single unwritten regional norm. They were versioned documents with their own architecture. That point can be established without turning this inquiry into a survey of three registry systems.

For ARIN, the active archived Number Resource Policy Manual immediately before AFRINIC-2 was version 2004.2, published on 16 December 2004. It contained a distinct section 4.3 for direct assignments to end users. For APNIC, the relevant archived comparison identifies APNIC-086 version 005, dated 16 August 2004, as the predecessor of version 006 dated 13 December 2005. The earlier document set out APNIC’s own IPv4 practices and policies, with its own assignment and small-multihoming structure.

In the RIPE NCC service region, RIPE-324, published on 28 October 2004, documented then-current IPv4 allocation and assignment policy, including utilization and provider-independent request materials.

The compact lesson is provenance. “Use the former RIR policy” was not operationally complete unless staff could say which named document and which version. A living policy corpus changes. A later copy may contain wording that did not govern on 27 April 2005. A change log may reveal that a test, definition or administrative form moved after the relevant date. Without version provenance, a reviewer cannot distinguish faithful continuity from retrospective substitution.

The differences also mattered to applicants. The three corpora varied in structure, eligibility language, assignment thresholds and administrative evidence. As an institutional inference, applying different former-RIR rules could impose different documentary burdens and could potentially yield different eligibility results for otherwise comparable requests. An applicant accustomed to one set of forms, definitions and multihoming expectations might face a different path from an applicant whose history lay with another registry.

That inference must remain bounded. There is no surviving request-level dataset here. No applicant is identified. No prefix is identified. There is no recorded acceptance, refusal, delay, appeal or quantified loss under the interim instruction. It would therefore be wrong to claim observed discrimination, to invent a representative case, or to attach a monetary figure to the gap. The policy texts show the possibility of differing burdens; they do not show who encountered them or what decision followed.

This evidentiary discipline is particularly important because the phrase “former RIR” can make historical fragmentation look neutral. History is a fact about a relationship, not automatically a justification for every consequence that follows from it. If two organisations seeking the same class of resource face different evidence demands solely because different registries previously served them, the coordinator should at least identify that difference, explain its temporary basis and provide a route for resolving anomalies. Otherwise, reliance becomes opaque path dependence.

The strongest case for borrowing the rule

The best defence of the 27 April instruction begins with the queue rather than with institutional prestige. A newly operational registry had to process requests. Pausing every end-user PI application until a complete local process ended could have left operators uncertain for months. Provider-independent address space can matter to an organisation’s ability to design a portable network relationship. Even without evidence of a specific stranded applicant, a general suspension would have transferred the cost of institutional transition to requestors.

Nor was case-by-case improvisation an attractive answer. If each officer invented a standard for each file, applicants would not know what evidence to prepare, similar files might receive inconsistent treatment, and later reviewers would struggle to reconstruct the basis of decisions. A published predecessor rule was more predictable and auditable than an unwritten exercise of staff preference.

Reliance adds a second reason. Members moving from an incumbent registry relationship could reasonably expect that an administrative transition would not instantly rewrite the eligibility framework for a pending or near-term request. Continuing the immediately preceding ruleset reduced the risk that the change of ledger operator itself would become the reason a request was treated differently. In that limited sense, predecessor continuity protected the user from the institution’s transition.

The alternatives each carried a cost. A total pause avoided applying a rule before local adoption but created service interruption. One uniform predecessor rule would have made all interim applicants face the same text, but choosing ARIN, APNIC or RIPE NCC as the universal temporary standard would itself have demanded a justification and could have erased legitimate reliance on prior treatment. Unbounded staff discretion offered flexibility for unusual cases but weakened predictability, comparability and auditability. Narrow predecessor continuation was therefore a defensible operational choice.

The defence is strong because it is narrow. It does not depend on saying that the predecessor registries had sovereign power that AFRINIC inherited. It does not depend on treating the CEO’s advice as legislation. It depends on a thin coordination objective: keep the ledger service functioning, preserve uniqueness and accurate records, and avoid arbitrary interruption while a local replacement is completed.

That same logic supplies the limit. Continuity does not answer which predecessor governs an ambiguous history. It does not select a policy version. It does not explain whether unrelated provisions travelled with the end-user rule. It does not tell an applicant why a particular evidentiary burden was chosen, what could cure a deficiency, whether a decision would be held while disputed, who could review it, or when the temporary instrument ceased to operate. Continuity can justify the bridge. It cannot justify leaving the bridge without rails, signs or an identifiable far bank.

A bookkeeper’s boundary

AFRINIC is a private ledger operator, bookkeeper and technical coordinator. Its legitimate common functions are thin: maintaining uniqueness, keeping accurate registry records, checking proof of control, recording security assertions, registering changes and transfers, preserving auditability, isolating disputes and maintaining operational continuity. It has no sovereign, legislative, police, punitive, confiscatory or transnational regulatory authority.

This boundary is not semantic housekeeping. It determines how the 2005 instruction should be read. A bookkeeper may need a temporary rule to decide what documentation accompanies an entry. It may need to know whether the applicant qualifies for the requested registry service. It may maintain a stable record while a dispute is separated for review. Those tasks are coordination. By contrast, the bookkeeper does not acquire ownership of the resource recorded, power to punish an operator for disfavoured conduct, or jurisdiction to legislate for a continent merely because other coordinators recognise its database.

The institutional form is consistent with that reading. Regional Internet registries are described as not-for-profit, member-based associations operating under the law of their respective countries, and AFRINIC is recorded as the fifth RIR, incorporated in April 2005. Membership and incorporation explain an association and a service framework. They do not create a sovereign.

The same discipline applies to procedural legitimacy. Open discussion, consensus, last call and Board ratification can make a private service rule more visible, contestable and stable. They can reduce surprise. They can help the coordinator document why it chose a common method. But procedure cannot manufacture powers the institution does not possess. Calling a meeting “the community” does not turn participants into a legislature for every affected network. Recognition by IANA or peer registries can coordinate a unique global ledger; it does not substitute for authority granted by the principals whose operations and resources are affected.

The risk on the other side is mandate laundering. A narrow need for record continuity can pass through compliance routines, interpretation and selective recognition until the institution begins to behave as though an inherited operational convenience were a permanent governing mandate. The movement can be gradual: first the registry selects a document, then it interprets ambiguous eligibility, then it treats deviation as non-compliance, and eventually administrative control is narrated as authority over the resource itself. The safeguard is to keep the borrowed rule attributable, versioned, temporary and reviewable.

Language should track the same boundary. If the meeting report says AFRINIC would “enforce” a policy, the safe reading is that staff would apply eligibility and evidence tests in processing the request. It does not support police power, punishment or confiscation. If official records speak of “authority” or “custody” in a handoff, the terms describe the institutions’ own coordination vocabulary. They do not prove title or public jurisdiction. Practical influence over service can be consequential without being sovereign.

The controls the short record does not show

A defensible interim instrument would have needed five connected controls: a crosswalk, provenance, reasons, review and a sunset.

The crosswalk would name the relevant predecessor provision for end-user PI requests and exclude unrelated policy fields. It would make clear whether the borrowed minimum size, utilization test, multihoming condition, evidence form and any exception travelled together or separately. It would also identify conflicts with AFRINIC’s existing operational forms. The purpose would not be to synthesize a new policy through staff interpretation. It would be to disclose exactly what the continuity instruction meant.

Version provenance would attach a date and document identity to each borrowed rule. For a request linked to ARIN, the file should show whether version 2004.2 supplied the applicable text. For an APNIC-linked member, it should preserve version 005 rather than allowing a later version to slide backward into the decision. For a RIPE NCC-linked member, it should identify RIPE-324 where that was the contemporaneous corpus. The record should survive even after online pages change, because a later transfer, audit or dispute may turn on the rule that applied when the registry entry was made.

Reasons would connect the selector to the request. A decision record should identify the former relationship relied on, the exact provision applied, the evidence considered and the remaining deficiency, if any. Notice would let the applicant know the rule before the consequential decision rather than discovering it only after refusal. A cure period would allow missing documentation or an identity mismatch to be corrected without converting an administrative defect into permanent exclusion.

Review would separate disagreement from punishment. A consequential registry decision should offer notice, reasons, an opportunity to cure, an appropriate stay and a route of appeal or independent reconsideration. A stay is especially important where an unresolved identity or version question could otherwise disrupt an operator while the institution decides which inherited rule it meant to use. Dispute isolation preserves the stable ledger state without pretending that the coordinator has police power.

Finally, the instruction needed a sunset tied to evidence. “Until this policy is adopted by AFRINIC” names a condition, but a usable cutover requires more. It should identify the replacement local policy, the adoption steps completed, the event that made it operational, and the treatment of requests filed before the cutover but still pending afterward. It should preserve the predecessor version and reasons for every concluded interim decision.

A temporary rule should not continue by habit after the replacement becomes operative, and the replacement should not retroactively reopen concluded interim decisions absent fraud, a duplicate claim or another narrow defect in ledger integrity.

The surviving meeting report does not publish these controls. That absence should not be inflated into a claim that none existed in practice. It means the instruction, as publicly recorded, cannot itself answer the questions. The proper response to a documentary gap is not to grant the institution the benefit of imagined powers. It is to state what remains unproved and identify the records that would resolve it.

A sunset visible only in stages

The later local policy can be identified, but its exact operational cutover cannot. The archive names AFPUB-2006-GEN-001, previously referenced as afpol-v4eu200504, “Policy for End User Assignments in AFRINIC service region,” credited to Mark Tinka. Its substance preserved the /24 minimum and the utilization tests discussed in Maputo.

The chronology attached to that archive requires care. The page carries a date of 16 May 2005. Yet its implementation history says consensus was reached at AFRINIC-3 on 13 December 2005, and that a 15-day last call ran from 9 to 24 February 2006. The page date therefore cannot safely be treated as the date of adoption, Board ratification, implementation or sunset of the borrowed rule.

There is a later outer boundary. At AFRINIC-4 on 17 May 2006, staff said AFRINIC had adopted the PI direct end-user assignments policy, and a Board member reported Board approval. That establishes that the local policy was described as adopted by that date. It does not establish the precise moment of Board ratification, the implementation timestamp used by staff, or the exact moment when pending files changed from predecessor rules to the local rule.

The documentary conclusion must remain deliberately modest. The interim instruction was designed to last until local adoption. Consensus was recorded in December 2005, last call in February 2006, and adoption and Board approval were reported by 17 May 2006. Somewhere across that sequence, a proper operational cutover should have ended the borrowed regime. The available records do not justify selecting an exact instant.

That uncertainty matters most for files near the boundary. A request submitted before cutover but decided after it might plausibly be treated under the filing-date rule, the decision-date rule or an expressly published transition rule. The meeting minutes do not choose among those approaches. It would be unsafe to infer one. The responsible design is prospective: publish the cutover, say how pending requests will be treated, preserve the rule applied to each file, and prevent the provisional instruction from lingering as institutional habit.

What can and cannot be concluded

The evidence supports a precise institutional conclusion. On 27 April 2005, in response to a question about end-user PI requests before the next member meeting, the CEO gave interim operating advice to use the policy of the member’s respective former RIR until AFRINIC adopted its own. The proposal under discussion had not completed the locally published process. The selector referred to historical registry affiliation. The predecessor corpora were distinct and versioned. The public record did not supply a crosswalk, edge-case method, reasons, review path or precise sunset.

The evidence does not support a story of proven abuse. No surviving Board resolution expressly authorising the interim instruction is identified. No implementation memorandum explains how staff interpreted “respective former RIR.” No request ledger shows the applicants, evidence, processing time or outcomes. No appeal is documented. No quantified cost or delay can be assigned. No observed disparate treatment can be claimed. The record also does not establish the exact Board-ratification or implementation date of the replacement policy.

Later economic conditions explain why the design question remains important, not why the participants acted in 2005. Different evidence and eligibility tests can affect staff time, applicant preparation costs, delay risk and the predictability of obtaining portable address space. An unpublished selector makes administrative risk harder to price and can favour organisations with greater legal, policy or documentation capacity. A dated sunset can reduce uncertainty in later transfers, financing, audits and disputes about how a registry entry originated.

Those are mechanisms, not historical outcomes. Nothing here attributes later IPv4 scarcity, liquidity concerns, market behaviour or IPv6 strategy to the people in the Maputo room. IPv6 does not answer an IPv4 PI governance gap, and this analysis is not an argument about IPv6 deployment. It is an examination of what a private coordinator may do when the local rule is unfinished and requests still need a predictable administrative path.

The answer is exacting but practical. Borrowing can be legitimate when it prevents interruption and preserves reliance. It remains legitimate only while it is narrow, attributed and temporary. The registry should show the predecessor relationship, the document version, the relevant provision, the reasons for the decision, the means of curing or reviewing an error, and the event that ends the temporary regime. Continuity is a service objective, not a source of sovereignty.

The 27 April instruction is therefore best understood as a bridge across a procedural gap. Its strength was that it avoided both a service freeze and improvised standards. Its weakness was that the surviving record did not specify the bridge’s lane markings or cutover. The proper institutional lesson is not that every interim act is invalid, nor that operational necessity grants unlimited discretion. It is that a ledger can remain open while its locally adopted rule is unfinished, provided the temporary rule is traceable and cannot become permanent by inertia.