Summary

  • The IPv6 Provider-Independent Update published on 28 March 2018 was AFPUB-2018-V6-004-DRAFT01. AFPUB-2018-V6-003 was a different proposal, the IPv6 Initial Allocation Update for LIR allocation criteria. Keeping those instruments separate is essential to understanding what actually changed.
  • Draft 1 attacked an obsolete dependency: an end-user organisation should not have needed an IPv4 PI past in order to establish a case for an independent IPv6 identity. It combined that opening with membership, need, site, sizing, deployment and aggregation conditions, plus a route for correcting an initial assignment that no longer fitted the network plan.
  • The legitimate case for those conditions is narrow but strong. A registry must avoid duplicate assignments, keep accountable records, size prefixes coherently and make routing consequences visible. None of those tasks gives a private institution jurisdiction to approve business models, compel global announcement, punish a routing choice or convert a database entry into a revocable favour.
  • Portability shifts a practical balance of power. Provider-assigned space binds continuity to the upstream whose aggregate contains it; PI space can let an operator change transit or multihome without replacing its public addresses. That benefit carries a routing cost, but the answer is proportionate evidence and transparent aggregation guidance, not discretionary control over whether an end user deserves an exit.
  • The durable rule is therefore thin: verify identity, control, uniqueness, sites and an objectively justified prefix size; record the result accurately; disclose operational implications; and provide correction and appeal. Questions of punishment, confiscation or adjudication belong to competent public authority, not the technical bookkeeper.

L3 — The IPv6 door that no longer required an IPv4 past

The first discipline in reading the March 2018 act is to identify it correctly. The IPv6 PI Update was published as AFPUB-2018-V6-004-DRAFT01 on 28 March 2018. The similar-looking identifier AFPUB-2018-V6-003 belongs to the separate IPv6 Initial Allocation Update. Both appeared in the same period, and both concerned IPv6, but they addressed different applicants and different control surfaces. V6-003 concerned initial and subsequent allocations for Local Internet Registries. V6-004 concerned provider-independent assignments for end-user organisations. Substituting one number for the other does not create a harmless citation blemish; it moves the analysis from an end user's route to stable addressing into the allocation architecture for organisations that distribute address space to customers.

The official archive, the historical policy list and AFRINIC's later annual account all preserve the distinction. Their evidential role is concrete and limited. They establish what the institution called the proposals, when it recorded them and how it later described their progress. They do not establish that a publication was already operative policy, still less that an entry in a private registry carried sovereign force. The exact institutional act on 28 March was publication of Draft 1. A public proposal had entered the policy discussion.

It had not yet travelled through the later meeting, board and implementation steps that AFRINIC itself recorded.

That chronology matters because several events are easily collapsed into one. At AFRINIC-28 on 6 June 2018, the proposal author explained the defect the draft was designed to address. The previous arrangement tied an applicant's route to IPv6 PI to the existence or qualification of IPv4 PI space. By then, using an IPv4 history as the ticket to an IPv6 resource was difficult to defend on its own terms: IPv4 exhaustion made the inherited condition less available precisely as organisations needed a practical way into IPv6. The meeting record says the proposal would allow an applicant without IPv4 PI space to seek IPv6 PI.

Participants questioned the twelve-month announcement condition, the continued role of the host-density ratio and the possibility of routing monitoring. The author said the host-density text remained and that staff were not being instructed to monitor routing.

The co-chairs moved V6-004 to last call despite one recorded objection. The objection was not frivolous. It favoured retaining the IPv4 qualification until suitable IPv6 qualification criteria had been defined. That concern exposes the genuine drafting problem: removing a poor proxy does not eliminate the need for an intelligible technical test. Yet the meeting record says there was no movement towards a change in the text, and the proposal advanced. Last call was a procedural stage after publication.

It was not publication itself, was not board ratification, did not implement the draft and did not transform a private coordination process into legislation.

The next institutional step came in August. Board minutes record Resolution 201808.449 as ratifying AFPUB-2018-V6-004-DRAFT01 on 7–8 August 2018 and directing staff to implement it. In November, the revision history associated the changes with sections 6.8 through 6.8.4, and AFRINIC announced partial implementation on 29 November, while saying that specified automation remained pending. These records make the sequence unusually clear: March publication, June last call, August ratification, November implementation work. They also prevent a misleading statement that the draft was fully operative on the day it was published.

Each official document proves an institutional act at its own date; none moves the effect of a later act backwards through time.

The substance of the opening was an end-user route. An organisation seeking it was not applying as an LIR for a stock from which to make customer assignments. It was seeking addressing for its own network as a non-LIR end user. The policy architecture required such an applicant to enter or hold an AFRINIC end-user membership relationship, identify itself, justify its number of end sites and demonstrate need. It also expected deployment of the requested space within twelve months. These conditions framed PI as a specific operational instrument rather than an undifferentiated reserve of addresses.

The sizing architecture attempted to translate a network plan into an administrable prefix. The later consolidated policy text and the later clarification preserve the relevant concepts, although neither should be mistaken for a verbatim copy of every word in March Draft 1. The architecture used at least a /48 per end site. For an organisation with several connected sites, the resulting prefix was to be aligned on a nibble boundary. A request for a shorter, and therefore larger, prefix required justification.

Site count informed the calculation; contiguous growth was preferred where possible; sparse allocation was part of AFRINIC's recordkeeping approach. These rules are important not because hexadecimal neatness has moral value, but because predictable boundaries can make registration, filtering, aggregation and future growth easier to understand.

Draft 1 also included a rectification concept for an initial IPv6 PI assignment whose original size no longer matched the later network plan. The meeting record identifies this as a deliberate addition. Rectification is a practical acknowledgment that forecasts fail. An organisation may discover that its real site structure or addressing requirements differ from the case made at the beginning. A correction mechanism can reduce the risk that an early sizing error becomes permanent fragmentation or forces the operator to improvise around a prefix that cannot coherently contain its legitimate use.

But rectification has two possible institutional characters. In its narrow form, it lets the record and assigned range be corrected so that a demonstrable network need can be represented coherently. That can protect continuity, particularly if contiguous expansion or a carefully managed transition is possible. In its broad form, it could become a claimed entitlement to force renumbering whenever staff later prefer another design. The evidence supports the existence of a rectification path; it does not supply an unlimited mandate to redesign networks.

The difference is the same throughout this policy: a technical device is legitimate to the extent that it protects a defined coordination invariant, and suspect when it becomes open-ended control over the operator.

Aggregation is another part of that architecture. Where PI space was announced, the policy sought aggregation as far as practicable so that avoidable growth in the global routing table could be minimised. That is a recognisable technical interest. A prefix independently visible in global routing may add state that routers and operators must carry. Coherent sizing and restrained deaggregation can make that cost more legible. Yet the wording must be handled carefully. An aggregation expectation for space that is announced is not the same thing as a rule that every legitimate PI use must be announced globally.

Nor does an assignment by AFRINIC guarantee that every network in the world will accept, propagate or route it. Registration, route origination, filtering and reachability are connected but distinct facts.

The later 2019 IPv6 PI Clarification illuminates why that distinction was not merely theoretical. It said the V6-004 revision had retained an announcement-and-reclamation sentence that did not fit valid cases in which PI space need not be globally announced, giving IXP peering LANs and private networks as examples. That later proposal is relevant as evidence of a drafting consequence noticed after V6-004. It must not be backdated into the March text as though the later replacement words were already there. It also provides no basis for claiming that AFRINIC actually reclaimed a named organisation's space.

No such named outcome is established in the record used here.

This event reconstruction therefore yields a more exact picture than either triumphal or hostile shorthand permits. V6-004 Draft 1 removed an inherited dependency on IPv4 PI status, set out a distinct end-user path, carried forward evidence and deployment conditions, described a sizing and aggregation architecture, and allowed initial-size rectification. It then advanced through later institutional stages. The opening mattered because it recognised an IPv6 need on its own merits.

The remaining language mattered because access to independent network identity can be hollow if the recordkeeper is allowed to turn operational preferences into a continuing licence.

There are also important things the documentary record does not establish. It does not quantify how many additional global routes resulted from V6-004. It does not identify a particular applicant that gained or lost service under the proposal. It does not document a named denial, reclamation or forced renumbering. It does not prove that every assigned PI prefix was universally routable. Those gaps are not weaknesses to be filled with supposition. They set the boundary between reconstruction and analysis: the institutional acts can be dated and described, while the policy's likely mechanisms can be tested without inventing outcomes.

The cleanest reading of the March act is consequently modest. AFRINIC published a proposed change to the criteria by which it would record a provider-independent IPv6 assignment for an end user in its service region. The proposal could make a meaningful difference to an organisation that previously lacked the requisite IPv4 history. It did not confer title in the sovereign sense, grant a public licence or create jurisdiction over the resulting network. The registry could coordinate a unique record. The operator, its upstreams and the wider routing system would still determine whether and how that prefix functioned in practice.