Summary

  • On 27 January 2009, AFRINIC's archive dates the first preserved text of AFPUB-2009-v6-001, “IPv6 Allocations to Non-Profit Networks.” Graham Beneke of the Johannesburg Area Wireless Users Group was the named author. The proposal offered a legally recognised not-for-profit responsible for free or low-cost network service in a local geographical area a special route to one provider-independent IPv6 /48. The organisation could qualify without planning a global announcement, but it was expected to describe peering with at least two public autonomous systems, announce the block in aggregate if it announced it, and refrain from sub-allocation.
  • The technical route was unusually bounded. An applicant could not already be an LIR or hold globally unique IPv4 space. Additional space would require a fresh justified request after subnet utilisation reached HD-ratio 0.94. Those conditions tried to distinguish an operating local network from an address reseller and to avoid casual sizing discretion. Yet the same text requested a special membership category and an “appropriate” fee structure without specifying a price, discount, payer, budget, cap, duration, review, audit or appeal.
  • That combination is why the proposal deserves close reading. For a genuine network with little or no revenue, a fixed administrative charge can be a serious barrier. But hiding a subsidy inside resource eligibility shifts costs and social choices without showing who bears them. AFRINIC is a private bookkeeper and coordinator, not a sovereign, regulator, welfare legislature or owner. The defensible design is therefore separation: retain transparent technical conditions for the /48 path, while placing any discount in an independently authorised, costed, published, capped, time-limited, auditable and reviewable fee schedule or grant mechanism.

The act was publication, not allocation. AFRINIC preserved a proposed rule that would have opened a narrow service path; the record does not show that a prefix changed hands, a fee changed, a beneficiary connected, or any deployment occurred. That distinction matters because a proposal can still expose an institutional choice before it produces an operational result. Here, the choice was whether technical coordination and social subsidy should occupy the same instrument.

The immediate stakes ran in two directions. A local non-profit network could have a large internal topology and valuable local peering while lacking either commercial revenue or a plan to announce globally. A rule designed around globally visible, conventionally financed applicants might therefore exclude a real operating network for reasons unrelated to uniqueness. At the same time, a private registry that verifies technical facts can slip into deciding social worth if its resource policy silently determines which organisational forms deserve cheaper service and who must absorb the difference.

The proposal's clauses allow those two problems to be separated rather than treated as a single moral question.

Two official dates, one unresolved relationship

The clearest event marker is 27 January 2009: that is the date displayed by AFRINIC's archive for AFPUB-2009-v6-001. The archived page names Graham Beneke and associates him with JAWUG. An official policy-summary deck for AFRINIC-10, however, displays 13 January 2009 beside the author and proposal. Both dates belong to official material, but the available record does not explain what each date represents.

It would be tempting to call 13 January the submission date and 27 January the publication date. The sources do not justify that neat sequence. Nor do they establish that 27 January was the proposal's first public appearance in every channel. The responsible account is narrower: the archive attaches 27 January to the first preserved proposal text, while a meeting summary carries 13 January, and the relationship between those labels remains unresolved.

Official meeting material places the proposal on the public-policy meeting agenda for 20 and 21 May 2009. That establishes entry into a formal private discussion process. It does not establish a result, a binding rule, implementation, or an allocation. A scheduled discussion is evidence that a text reached an institutional forum; it is not evidence that participants possessed public-law authority or that the proposal took effect.

This date discipline is more than archival housekeeping. If a reader is allowed to mistake a draft for an operative rule, every later inference becomes inflated: an offered /48 turns into a delivered /48, a requested fee category turns into an adopted subsidy, and discussion turns into mandate. The record supports none of those transformations. The event is important precisely because the first text shows the intended mechanism in a reversible, pre-operational form.

The beneficiary was defined by service, place and legal responsibility

The proposed beneficiary was not “the community” in the abstract. It was an organisation legally responsible for a network providing free or low-cost services within a local geographical area, and that organisation had to be recognised as not-for-profit under the law relevant to its country or region. Each part of that description did different work.

The service element aimed at a network offered free or at low cost. The geographical element bounded the operating claim to a local area. The legal element placed responsibility on an identifiable organisation rather than an informal collection of devices or participants. And the not-for-profit test relied on recognition supplied by relevant law, not on AFRINIC inventing a new civil status.

This architecture has a legitimate administrative core. A private coordinator can ask an applicant to prove who it is, who is responsible for the network, and whether it fits a published service category. It can verify documents for the limited purpose of deciding access to that service. Domestic legal recognition remains the source of the not-for-profit status; AFRINIC would record and check evidence rather than legislate the status into existence.

The first text, however, left the verification boundary unfinished. It did not publish a document checklist, explain how an organisation recognised in one jurisdiction would be treated in another, resolve conflict-of-law questions, establish a cadence for rechecking status, or specify a correction or appeal path. Those omissions do not make the beneficiary definition meaningless. They show where an apparently objective label could generate staff discretion once real applications presented mixed legal forms.

Legal form is also an imperfect proxy for economic need. A not-for-profit may have strong funding, while an informal or commercially organised local network may have little cash. Recognition under law can make administration more predictable, but it cannot by itself establish inability to pay, technical readiness, public value, or the scale of the service. The proposal used legal status as a gate; it did not prove that the gate measured every quality its fiscal ambition seemed to assume.

The text's references to free service, low-cost service and networks without ordinary revenue should therefore remain claims embedded in the proposal's rationale, not findings about an entire population. There is no verified list of candidate networks, no proven count, no financial survey, and no documented topology for a named applicant in the available record. A plausible beneficiary class is not the same thing as a measured beneficiary population.

Six controls made the route narrow

Once the beneficiary definition was satisfied, the proposal placed several distinct controls around the special path. They should not be blended, because each addresses a different risk.

First, the applicant could not already be an LIR. That condition separated the path from the ordinary position of an organisation already receiving registry service in that capacity. Non-eligibility would not have been a punishment; it would merely mean that the organisation did not qualify for this particular offer.

Second, the applicant could not already hold a globally unique IPv4 assignment. The likely targeting logic is apparent: reserve the special route for networks lacking that form of globally unique address foothold. Yet the condition is blunt. Possession of an IPv4 assignment could indicate earlier success, operational capacity or a legacy need; it does not necessarily prove present ability to finance IPv6 service. The rule reduced ambiguity while risking exclusion of technically prepared networks.

Third, the initial size was fixed at one IPv6 PI /48. A fixed size limits negotiation and removes pressure for staff to make case-by-case judgments about a larger or smaller initial block. The prefix was provider-independent in the proposed text, but the offer of that administrative category did not guarantee that another network would accept a route, that transit would be available, or that the applicant would achieve reachability.

Fourth, an applicant did not need a plan to announce the prefix globally. This clause recognised that local peering can be an operating purpose in its own right. It prevented global visibility from becoming a proxy for legitimacy. At the same time, the absence of a global-announcement requirement did not prove that a local route would be accepted, exchanged, secured or useful. It removed one eligibility barrier; it did not manufacture connectivity.

Fifth, the applicant should provide details of peering with at least two public autonomous systems. The word “should” matters. The text did not frame the evidence in the same compulsory terms as every other condition, and it did not define what documentation would suffice. Two public-AS peerings would indicate an interconnection surface, but not necessarily two upstream providers, paid transit contracts, resilient multihoming, global reachability, useful traffic exchange, route acceptance or deployment. Treating the number two as a complete quality certificate would say much more than the proposal did.

Sixth, sub-allocation was prohibited. This was an anti-arbitrage boundary: the recipient was not to turn a discounted or specially eligible block into an onward address-distribution business. It helped distinguish use by the responsible network from operation as a shadow provider. But no sub-allocation did not mean no internal subnetting. It did not prohibit devices, links, internal segments or multiple /64 networks inside the recipient's own topology.

Together, these controls formed a recognisable screening stack. Legal responsibility identified the accountable organisation. The LIR and IPv4 exclusions targeted a special entrant class. The one-/48 limit bounded the initial resource. The peering evidence connected the claim to an operating network. Aggregation protected routing tidiness. No sub-allocation reduced resale risk. A later expansion threshold discouraged casual return for more space. The design was not merely an appeal to sympathy.

Yet bounded does not mean complete. The proposal did not specify how contradictory evidence would be resolved, whether “should” functioned as guidance or a practical threshold, or which failures could be corrected. The special path could be narrow without being arbitrary only if its evidence rules were public, consistently applied and reviewable.

What one /48 meant inside the network

The one-/48 offer carried more internal capacity than a casual reading of “one block” might suggest. Under the ordinary /64 subnet boundary used for the relevant IPv6 addressing model, the difference between /48 and /64 leaves 16 subnet bits. Sixteen bits yield 65,536 possible /64 subnets within the /48.

That arithmetic explains why a fixed /48 could serve a network with many internal segments without repeated initial sizing disputes. The recipient could structure its own topology with multiple /64s while remaining within one aggregated prefix. The proposal's no-sub-allocation clause targeted onward distribution, not the organisation's internal design. Confusing those concepts would turn an anti-resale rule into a fictional ban on operating a network.

The /48 was also a contemporaneously recognisable site-planning size. That context helps explain why the proposal chose a fixed block rather than an exotic number. It does not prove that every eligible organisation was a single site, that each needed all possible subnets, or that the prefix would be globally visible. It establishes technical plausibility, not measured demand.

A globally unique prefix can reduce address ambiguity and support direct reachability where routing and security policies permit. It cannot, by itself, buy transit, install radios or routers, configure filters, register routes, secure route origin assertions, create traffic demand, or persuade peers to accept an announcement. The proposal addressed an address-coordination barrier. It did not supply the rest of an operating network.

The technical context also clarifies the limits of purely local address alternatives invoked by the proposal's background. Private IPv4 space carries external-connectivity limits, and IPv6 unique local addresses are intended for local communication rather than expected global routing. Those facts make a globally unique /48 intelligible for a network that values stable interconnection. They do not prove that a named applicant suffered a documented failure, that global transit was unobtainable, or that every local network needed provider-independent space.

Optional global reach, conditional aggregation

The most interesting technical tension sits between two clauses. An organisation could qualify without planning to announce the prefix globally. Yet the proposal also said that the allocation must be announced in an aggregated way. Read literally and without context, those statements can look inconsistent: how can announcement be mandatory if no announcement plan is required?

The most coherent reading is conditional. If the organisation announces the prefix, it should announce the /48 as one aggregate rather than fragmenting it into more-specific routes. That interpretation protects routing aggregation while preserving a local-peering or non-global use case. But it remains an interpretation. The first text did not expressly distinguish private peering, local route exchange and global announcement, nor did it specify what happened when the recipient made no announcement at all.

This ambiguity matters operationally. A local network might exchange the aggregate with selected peers without seeking broad global propagation. Another might initially remain local and later seek wider reach. Filters applied by other operators could accept or reject the route independently of AFRINIC's record. The registry could maintain a unique record and a route-registration surface; it could not command the global routing system to carry the prefix.

Aggregation is a sensible coordination objective because one route is easier to reason about than numerous fragments. It is not an enforcement licence. AFRINIC's role is to administer transparent service terms and accurate records, not to police the Internet, prosecute routing conduct or confiscate resources. In this first instrument, aggregation was a condition attached to a proposed service path for unallocated inventory, not proof of coercive authority over existing networks.

The peering clause should be read with the same restraint. Details concerning at least two public ASes could demonstrate that the applicant was not describing a purely imaginary topology. Still, the text did not name the ASes, require contracts, measure traffic, test capacity, define “public,” or show route acceptance. Evidence of relationship is not evidence of result. The applicant would control its network design and peering choices; AFRINIC would record eligibility, not operate those links.

HD-ratio 0.94 was not 94 per cent occupancy

The proposal limited the first grant to one /48, then required a fresh justified application for additional resources after subnet utilisation reached HD-ratio 0.94. This was a density threshold, not permission to expand automatically. The words “fresh” and “justified” matter: reaching a numerical threshold would open a new request, not silently enlarge the original block.

HD-ratio is logarithmic. In simplified form, it compares the logarithm of used objects with the logarithm of the maximum possible objects. It does not mean that 0.94 equals 94 per cent flat occupancy. Treating it that way would materially misstate the expansion control.

The proposal did not identify the object counted for this calculation or the rounding method. A transparent illustration is possible only by stating an assumption. If /64 subnets are treated as the objects, one /48 contains 65,536 possible objects. Raising 65,536 to the power 0.94 gives approximately 33,689.23, conventionally rounded to about 33,690 used /64 objects. That is roughly 51.4 per cent of the 65,536 possible /64s.

The result is an inference, not a number printed in the proposal. A different object definition or rounding convention could change the operational threshold. Before administering such a rule, a responsible service would have to publish the counted unit, acceptable utilisation evidence, calculation method and treatment of borderline values.

Even with that missing precision, the clause reveals a useful design instinct. It tied expansion to demonstrated density rather than institutional prestige or open-ended claims about future growth. It also preserved a separate justification step. In a narrow coordination system, measurable resource conditions are preferable to discretionary judgments about whether an applicant seems deserving. The metric becomes trustworthy, however, only when its unit and method are explicit.

The fiscal half was a request, not a design

The proposal's technical half was specific enough to reconstruct. Its fiscal half was not. The text called for a special membership category and an appropriate fee structure because it described many intended networks as free, low-cost or without revenue. But “appropriate” supplied neither a price nor a rule for arriving at one.

There was no proposed discount percentage, fixed charge, setup treatment, renewal price, subsidy amount, payer, appropriation, budget, beneficiary cap, aggregate cost, duration, verification cadence, audit, public report, review date, appeal or exit rule. There was no answer to whether the difference would be carried by other members, corporate reserves, a donor, lower service expenditure, or some combination. Zero specified subsidy dollars is not the same as zero fiscal effect; it means the incidence was hidden.

Contemporaneous corporate planning material offers only a narrow point of comparison. AFRINIC's 2009–2011 strategy-plan table printed USD 2,500 for setup and USD 2,500 for membership on its IPv6 row. The table does not establish a universal invoice, prove that both values would have been charged cumulatively to an applicant under the proposed path, define academic discounts, or state the special rate Beneke's proposal sought. Calling it a USD 5,000 first-year bill would convert two planning-table entries into an invoice the evidence does not provide.

Still, those printed values help explain why fee design could not be treated as decorative. A fixed amount falls more heavily on a small or no-revenue organisation because it is spread across fewer paying users or no ordinary income. Legal work, application effort and recurring administration can add burdens beyond the headline price. Affordability was therefore a plausible problem. The absence of a measured applicant population or verified finances means its scale cannot be quantified here.

The institutional problem begins when a plausible burden is translated directly into an uncosted preference. A lower price has an opportunity cost even when no cash leaves the registry. Someone forgoes revenue, finances administrative work, absorbs risk or accepts reduced service capacity. Without an identified payer, an apparent benefit to one applicant becomes an undisclosed liability elsewhere.

A private membership corporation may set transparent prices for its own service. It may establish a discount under proper corporate authority. A sponsor may pay a standard fee for an eligible organisation. None of those acts requires sovereign power. What a private policy discussion cannot do is transform an unspecified social aspiration into public fiscal authority merely by attaching it to number-resource criteria.

Fee and resource design answer different questions. Resource policy should ask whether a proposed assignment protects uniqueness, accurate records, aggregation and bounded use. Fee policy should ask what the service costs, who pays, who qualifies for relief, how much relief is available, how long it lasts, and how decisions can be reviewed. Combining the questions invites technical staff to make social-allocation judgments without a clear mandate or accounting trail.

The strongest defence of the proposal

The best case for AFPUB-2009-v6-001 is stronger than a caricature about free addresses. It identified a plausible mismatch between conventional eligibility and an operating local network. A community network might have many internal segments and meaningful peer relationships while lacking global transit and ordinary commercial revenue. Requiring a plan for broad global visibility could reject such a network even when globally unique addressing would improve local interconnection.

The proposed remedy constrained itself. One /48 was a clear ceiling for the initial request. Legal responsibility attached the network to a recognised organisation. Existing LIRs and holders of globally unique IPv4 space were excluded. Peering details offered evidence of an interconnection surface. Aggregate announcement discouraged fragmentation. No sub-allocation reduced the chance that a specially priced recipient would become an address reseller. HD-ratio 0.94 and a fresh justification controlled expansion.

Those clauses collectively tried to target use, limit arbitrage and keep routing orderly. They also reduced the temptation to administer charity case by case. If ordinary fees had been designed around commercial or academic institutions, a separate category could make a technically valid service available to a different organisational form. Lowering a fixed-cost barrier without granting scarce IPv4 space was a coherent goal within the proposal's own frame.

The defence is important because the correct criticism is not that a private coordinator must ignore affordability. Private actors price services constantly. Nor is the answer to insist on global announcement as the only evidence of a real network. A narrow local-peering route can be legitimate. The better question is which technical clauses protect the common coordination layer and which fiscal choices require a separate authority and ledger.

The strongest critique

The proposal's targeting rules could exclude worthy applicants and admit financially comfortable ones. Not-for-profit form does not measure cash need. A network with a globally unique IPv4 assignment might be especially capable of deploying IPv6, yet the exclusion would remove it from the special path. An organisation without such an assignment might still have substantial funding. Administrative simplicity came at the price of imperfect economic targeting.

Two public-AS peerings were also a weak proxy if treated as a hard proof of utility. They did not establish that routes would be accepted, that traffic would flow, that capacity was meaningful, that the network had two upstream providers, or that its connectivity was resilient. Because the wording said the applicant “should” provide details, an implementation would have needed to clarify whether missing evidence caused non-eligibility or merely invited explanation.

The no-global-announcement clause and aggregation clause required reconciliation before consistent administration. The HD ratio required an object definition and rounding rule. The legal-status test required a document and correction process. These were repairable gaps in a technical service design, but they mattered because every undefined term expands staff discretion.

The fiscal gap was more fundamental. The proposal gave no cost model and identified no payer. It did not estimate how many organisations might qualify, how a discount would affect corporate revenue, how often eligibility would be checked, or when the category would end. Without a cap and review, even a small concession can become an open-ended precedent. Without an appeal, applicants cannot distinguish a technical rejection from a legal-status dispute or a fee decision.

Resource policy should not decide, silently, who finances a social goal. If other members are expected to pay through higher charges or forgone service, that incidence should be visible. If reserves are used, the budget and limit should be published. If a donor pays, selection rules and continuity risk should be disclosed. If the standard fee remains, the technical exception should be described honestly as solving routing eligibility rather than affordability.

This critique does not deny the proposed beneficiaries' possible needs. It demands a mechanism equal to the claim. Sympathy is not an appropriation; “appropriate” is not a tariff; and participation in a policy meeting is not fiscal authority.

The bookkeeper boundary

AFRINIC's legitimate role in this design was useful and narrow. As a private, non-governmental, not-for-profit membership corporation, it could receive an application, verify documents for its service, maintain unique records, assign unallocated inventory under transparent rules, keep registration data and charge disclosed fees. It could publish a proposal and convene a discussion. Those are acts of private coordination.

AFRINIC was not a state, sovereign, legislature, regulator, police force, prosecutor, court, confiscator or owner of the networks and number resources it recorded. The proposal could not create those powers. A meeting could refine private service terms; it could not vote continental sovereignty into existence. A community could discuss and advise; participation did not turn it into a public principal.

This boundary does not imply institutional paralysis. A bookkeeper can make meaningful choices about the integrity of the book. Uniqueness, accurate records, proof of control, interoperability, routing-security assertions, auditable changes and replaceability are legitimate coordination concerns. Aggregate announcement and no sub-allocation can be evaluated against that thin common layer. A fixed /48 can reduce arbitrary sizing. Public evidence requirements can make service predictable.

The boundary becomes critical when the institution moves from verifying a legal fact to judging social worth. Checking that an applicant has recognised not-for-profit status for a disclosed category is a private administrative act. Deciding that one class deserves an undisclosed transfer from another, without a budget or review, is a fiscal allocation choice. The latter may still be undertaken by a private corporation under proper authority, but it must be named as pricing or subsidy—not disguised as the natural consequence of address policy.

Nor should eligibility be described as enforcement. If an LIR, an IPv4 holder or a commercial ISP did not fit the special route, the first text merely left it outside that proposed service path. It did not establish a penalty, seizure, prosecution or adverse judgment over existing holdings. The no-sub-allocation rule defined use of a prospective assignment; it did not prove police power.

The applicant, meanwhile, would remain responsible for its network. Its engineers and peers would determine topology, route exchange, security and service. AFRINIC's record could coordinate uniqueness and support accurate registration. It could not supply reachability by decree, own the network or control the decisions of independent operators.

The control chain therefore began with domestic law recognising the organisation, continued through applicant evidence and published technical conditions, and ended with private administrative recording. Any fee relief needed a second chain: corporate or external authorisation, an identified source of funds, published terms, accounting, review and an exit. The proposal described the first chain more clearly than the second.

What the record does not establish

No available evidence identifies an actual applicant, prefix, route, peer, allocation, denial, appeal, deployment, cost saving or abuse under this first text. There is no verified list of candidate networks, no audited revenue profile and no measured estimate of take-up. The proposal's descriptions of large device populations, licence-exempt radio use and absent revenue remain proposer background rather than independently established population facts.

The record does not show that an existing rule denied a named network, that transit was unavailable, or that local addressing caused a documented failure. It does not explain why two public ASes were selected, define acceptable peering evidence, resolve the announcement tension, or specify the HD-ratio object. It provides no original submission email, signed authorisation or pristine timestamped version history that resolves the two official dates.

There is also no direct event-specific material from NRS, Heng Lu, LARUS or BTW concerning AFPUB-2009-v6-001. Heng Lu's doctrine controls the authority analysis: private bookkeeping, uniqueness, visible prices and no sovereign or punitive power. It is not independent chronology for the January 2009 act. Official records establish what AFRINIC published, scheduled and printed; they do not by themselves prove fairness, representative mandate, social benefit or lawful public authority.

These absences discipline the conclusion. The first instrument was a serious attempt to lower an administrative addressing barrier. Its one-/48 path had a coherent anti-arbitrage structure, although several terms needed operational definition. Its request for a special fee category expressed a plausible affordability concern but did not design a subsidy. The durable answer is to preserve the technical path as transparent coordination and move financial relief into a separately authorised mechanism whose payer and limits can be seen.