Summary

  • In May 2006, AFRINIC's Board ratified a proposed change to the ordinary ASN assignment criteria under Resolution 200605.27 and instructed staff to implement it. The preserved comparison shows that the amendment added only a requirement for the organisation to be an AFRINIC member in good standing, of End-User or LIR type; the distinct-routing-policy, multi-homing, prospective-qualification and RFC 1930 clauses were already present.
  • The formal bullets do not print either AND or OR between unique routing policy and multi-homing. A July 2006 NRO comparison summarised AFRINIC's test with OR, but that table was not the adopted text and does not reveal how staff decided individual requests.
  • The applicant bore the initial burden of showing a need for a globally unique ASN and its good-standing service status, while staff could ask enough information to understand the planned routing policy. No 2006 form, evidence checklist, applicant correspondence, decision ledger, reasons code, cure process or review route has been recovered in the public chain.
  • The strongest institutional reading is narrow. AFRINIC was coordinating a unique ledger entry, not granting permission to run a network. It was a private bookkeeper and uniqueness coordinator, not a sovereign, regulator, licensor, police force, prosecutor, court, punishment authority or confiscatory power.

One new line, not a new routing test

The May 2006 Board entry is concise. After receiving a report from the Policy Working Group chair, the Board recorded its ratification of several open proposals, including the change to criteria for ASN assignments, and directed staff to implement them. Resolution 200605.27 is an identifier, not proof that the Board acted on 27 May. The official page groups the decision under May 2006 but supplies no exact calendar day. It also gives no individual votes, quorum record, attached final text, reasons or implementation checklist.

The claim supported by that page is therefore important but bounded: the Board ratified in May and handed implementation to staff.

The amendment itself was narrower than the resolution's institutional weight might suggest. The surviving “was” and “should be” comparison retains the technical eligibility language and inserts one requirement: the organisation must be an AFRINIC member in good standing, of End-User or LIR type. It did not invent a distinct-routing-policy test. It did not introduce multi-homing as a criterion. It did not create the route by which an organisation could show that it would qualify after assignment or within a reasonably short time. Nor did it newly subject eligibility to evaluation under RFC 1930.

All four elements were already in the text that Mark Tinka proposed to amend.

That distinction matters because it fixes both the event and the institution's claimed control surface. The ratified change connected technical eligibility to a current service relationship. It made good standing explicit in the assignment document, allowing the registry to identify the party asking it to write a globally unique number into its records. It did not broaden the underlying technical inquiry into a general power to approve business plans or network architecture.

Any analysis that attributes the whole routing test to Resolution 200605.27 overstates what changed and obscures the real issue: what the addition of account status did to a pre-existing test whose conjunction was already unclear.

What the retained technical propositions meant

The first retained proposition was a unique routing policy, parenthetically defined as one different from the policies of the organisation's border gateway peers. “Unique” in this context did not mean inventive, commercially distinctive or internally complex. It referred to externally visible routing choices that could not be represented faithfully as part of a provider's autonomous system. The relevant comparison was with the policies of adjacent external systems.

Merely exchanging routes, configuring BGP, deploying redundant routers or maintaining a sophisticated interior network did not by itself prove that the organisation needed a separate globally unique policy identity.

RFC 1930 helps interpret the distinction without turning technical advice into law. A single-homed site whose policy is the same as its provider's normally does not require its own globally unique ASN; the provider's AS can represent that relationship. A rare single-homed site may nevertheless have an externally meaningful policy sufficiently different from its provider's to justify a separate number. Conversely, an organisation can be technically active and resilient without establishing that a new globally unique ASN would add useful exterior-routing information.

Internal redundancy and external policy identity answer different questions.

The second retained proposition was that the site be multi-homed. Here the operative fact was connection to more than one other autonomous system, not simply multiple links that terminate in the same external system. A later consolidated AFRINIC manual also recognised connection to a public Internet exchange point. That later wording helps explain the retained administrative model, but it cannot be treated as a verbatim account of the evidence staff demanded in May 2006. The stable conceptual point is that multi-homing concerned multiple external routing relationships.

Two circuits, two routers or two entrances to one provider did not automatically establish that proposition.

The distinction between these propositions is not academic. An operator might have multiple external connections yet apply substantially the same policy towards them. Another might have only one current external connection but a genuinely distinct policy that must be expressed independently. RFC 1930's guidance makes room for those differences. It asks whether global routing gains meaningful policy information from a distinct number, rather than treating the existence of equipment or a protocol session as a sufficient entitlement.

That is why the exact relationship between the two bullets—alternative routes to eligibility or cumulative demands—changes the practical reach of the test.

The third retained element addressed future qualification. An organisation that did not yet satisfy the listed condition or conditions could qualify by demonstrating that it would do so after receiving the ASN or within a reasonably short time. This was a sensible escape from circularity: an operator could need the identifier before it could activate the proposed external arrangement. But the phrase “reasonably short time” remained unquantified in the surviving record.

There is no recovered 2006 rule stating how many days or months counted, or whether a forecast, planned activation date, peer confirmation, contract, configuration or some other document was required.

The fourth element directed evaluation through RFC 1930. Its disciplined use would focus attention on whether a globally unique number was technically useful, not on whether the registry approved of an operator's commercial strategy. The memo cautions that external routing activity and speculative future-proofing do not automatically justify a unique ASN. A private-use or provider ASN may be sufficient where global uniqueness and global announcement are unnecessary. Yet the memo is technical guidance. It does not confer public authority, and it does not establish how a particular AFRINIC staff member read a particular request in 2006.

The conjunction the formal text did not supply

The most consequential small omission lies between the first two bullets. The formal proposal lists a unique routing policy and a multi-homed site but prints no explicit AND or OR. Punctuation and sequence do not eliminate that ambiguity. A cumulative reading would require both a policy different from border peers and multi-homing. An alternative reading would allow either proposition to establish eligibility, subject to the surrounding necessity evaluation. The public text does not conclusively select one.

A contemporaneous NRO comparative overview, published in July 2006, described AFRINIC's eligibility test as unique routing policy OR multi-homing. That is meaningful evidence because it shows how the regional criterion was publicly summarised close to the ratification. It also fits the possibility, recognised in RFC 1930, that a rare single-homed site with a genuinely distinct external policy may need a separate ASN. But the comparison was not a formal AFRINIC amendment, a Board attachment, a staff instruction or an individual-decision record.

It cannot silently insert OR into the ratified text or prove that every request received that reading.

Nor would it be sound to impose AND simply because the proposal used separate bullets beneath an eligibility heading. The future-qualification clause referred back to “the criteria” without resolving whether that meant each bullet or an applicable limb. A later consolidated manual preserved the clauses but does not recover the missing conjunction as applied at the time of ratification.

The correct historical conclusion is therefore uncomfortable but precise: the formal language was ambiguous; the NRO summary used OR; and the public record does not show whether staff applied an alternative, cumulative or contextual test to individual 2006 requests.

This uncertainty would have affected what an applicant thought it had to prove. Under OR, evidence of genuine multi-homing could suffice even if the policies towards the external systems were not meaningfully different, while a rare distinct-policy case could qualify without current multi-homing. Under AND, each applicant would need to establish both propositions, potentially excluding the exceptional single-homed case that the technical guidance recognises. Under a contextual interpretation, staff discretion would carry even more weight. Without a reasons ledger, no reader can determine whether similar cases were treated alike.

From Mark Tinka's post to Board ratification

The record begins with a proposal dated 16 April 2005 and posted by Mark Tinka to the AFRINIC policy list on 17 April. Tinka, then identified with Africa Online Swaziland, proposed the wording. He did not enact it, report meeting consensus, ratify it on behalf of the Board or decide an ASN application. His stated rationale was to make the membership criterion explicit and identify organisations using ASNs through AFRINIC. Language about improving control should be read in that service-document context; it does not establish regulatory power.

The list exchange quickly exposed the practical meaning of the proposed line. Alan Levin said he was not yet in favour, and Gregory Massel objected to what he understood as an extra annual membership charge for an organisation holding only an ASN. These messages establish the concerns of named participants, not the size or balance of the constituency. They show that membership status was not viewed as empty administrative wording: participants connected it to price, category and whether a resource holder would have to pay an additional amount.

Adiel Akplogan, participating as AFRINIC's chief executive, answered that USD 400 was a setup fee. He later clarified that an ASN holder belonged to a distinct membership category costing USD 50 annually and distinguished it from a USD 400 member-only category for an organisation without resources. Those figures must remain attributed to his contemporaneous explanation. The messages are not an independently audited fee schedule, do not establish what every organisation ultimately paid and do not prove that the clarification resolved every objection.

Levin replied that the clarification helped, then asked whether the added line was redundant if the fee arrangements already established the membership relationship. Akplogan answered that the amendment completed the policy document because the fee document and the policy document served different functions. That response states AFRINIC's position on document design: a condition affecting eligibility should appear in the policy rather than remain inferable from a separate fee instrument. It does not by itself settle a legal characterisation or prove community-wide agreement.

It does, however, sharpen the proposal's true institutional effect. Good standing was being made an explicit issuance gate.

On 13 December 2005, the AFRINIC-3 meeting report listed the change to ASN assignment criteria among the presented proposals. It reported that the proposals received consensus and would return to the mailing list for a 15-day last call before submission to the Board. The report is evidence of the meeting process as officially recorded. It does not give a vote count, denominator, show of hands, transcript, dissent ledger, revised text or detailed method by which consensus was determined. “Consensus” is therefore a reported process outcome, not a licence to invent unanimity.

The official proposal archive records a last-call period from 9 to 24 February 2006. The page also displays “Date 17 May 2005”, a date that conflicts with the later reported meeting, last call and May 2006 ratification if treated as an implementation date. The responsible reading is to label it archive metadata whose exact meaning is unresolved. It may concern publication, migration or another archival event, but the available record does not say. It cannot be used to move implementation back to May 2005.

The chain then reaches Resolution 200605.27 in May 2006. The Board entry says that, following the working-group chair's report, the Board ratified four open proposals, including this one, and instructed staff to implement them. The grouped entry also covered other matters, but this report isolates the ordinary ASN eligibility amendment. Nothing in the identifier establishes the exact day. Nothing in the surviving entry supplies the operational notice that would show which final wording, conjunction or proof threshold staff received.

The later consolidated Resource Policy Manual closes part of the documentary chain. It preserved the eligibility clauses, the good-standing requirement and AFRINIC's ability to request information necessary to understand the planned routing policy and decide whether an ASN was needed. It also described public aut-num registration, administrative and technical contacts, recommended routing-policy registration, continuing update duties and the return of an unused ASN. These provisions show an administrative model built around assignment and record maintenance.

They do not reconstruct the exact May 2006 request form or prove how a hostmaster handled any application.

Membership was a service fact, not a fitness judgement

The good-standing line added a separate proposition to the decision. An applicant now had to be an AFRINIC member in good standing, of End-User or LIR type. On the most defensible reading, that status identified the counterparty to a private coordination service, connected the assignment to current contact and account records, and supplied a route for service costs and maintenance duties. It did not declare the organisation solvent, virtuous, competent, publicly fit or legally authorised to run a network.

That separation is essential because technical need and account status can fail for different reasons. An applicant may present a convincing case for a globally unique routing identity while having an incomplete or disputed service account. Another may be fully paid and administratively current while failing to show why a provider or private ASN is insufficient. If those findings collapse into one unexplained “not eligible” result, neither the applicant nor a later reviewer can tell whether the defect concerns routing facts or a curable account matter.

The mailing-list exchange shows why the distinction had economic force. Participants questioned whether the new wording meant another annual cost. AFRINIC's chief executive responded with different amounts and categories and argued that policy and fee documents had separate jobs. The most that can be concluded is that cost and category were contested, clarified in those terms, and then treated by AFRINIC as reasons to state membership expressly. It would be unfair to call the amendment a covert fee increase on that record. It would be equally incomplete to pretend that good standing had no price or bargaining implications.

Good standing also created an incentive for the registry to keep technical eligibility and the service relationship aligned. Accurate identity and contact data make a uniqueness ledger more useful, while a continuing relationship can support updates and the orderly return of an unused number. The danger begins when account status is treated as if it proves or disproves network necessity. A USD-denominated membership category cannot establish whether external policy is distinct. Conversely, an impressive routing plan cannot establish that a service account is current. An accountable decision would report both findings separately.

The applicant's burden and the limits of staff inquiry

The applicant bore the initial burden because the decisive facts concerned its planned external routing. At minimum, an intelligible claim would identify the requesting organisation and applicable membership category; explain the external routing policy and how it differed from border peers if relying on that limb; identify the external autonomous systems or public exchange connection that made the site multi-homed if relying on that limb; explain what would change after assignment if relying on prospective qualification; and show why a provider or private ASN would not represent the needed globally visible policy.

Those are propositions supported by the text and technical guidance, not a recovered list of documents. The public record does not show whether 2006 staff demanded peer contracts, letters, router configurations, policy statements, invoices, activation plans or any other particular item. It does not identify the evidential threshold or say how staff tested a planned relationship that did not yet exist. A modern observer may design a sensible application record around those propositions, but must not project that design backwards as a historical form.

The later manual supports a staff power to ask for enough information to understand planned routing policy and decide whether an ASN was needed. “Ask” matters. In a private service assessment, AFRINIC could make requested information a condition of completing its assignment decision. It did not possess regulatory discovery powers or legal compulsion over the applicant. The scope of a legitimate question came from the stated coordination proposition: identity, good standing, distinct external policy, multi-homing, credible near-term activation or the need for global uniqueness.

Questions untethered from those propositions would cross the line. The registry did not need to decide whether it preferred the applicant's business model, commercial counterparties, market entry, prices, corporate strategy or topology. Some commercial evidence might incidentally help verify a planned peer relationship, but that does not make commercial design the object of the decision. The proper inquiry is always, “What locally verifiable fact does this item establish about the need for a unique ledger record?” If no answer exists, the request for information is difficult to justify under the narrow service function.

The output was correspondingly limited. If satisfied, AFRINIC could assign a globally unique ASN and register it in the coordination system. If not satisfied, it could decline that ledger entry under its stated service criteria. A refusal did not declare the applicant's network illegal, forbid the organisation to operate, invalidate its contracts or command remote networks. The operator could still choose its topology and seek other technically available arrangements, while every remote autonomous system retained its independent choice to accept, filter or disregard routing information.

The strongest case for screening

The case for a meaningful gate deserves to be stated at full strength. A globally unique ASN is not ordinary stock whose value exists in isolation. It functions because operators can distinguish policy identities consistently across an interconnected environment. A needless number can introduce another global identity without adding useful routing information, creating small but distributed costs in configuration, documentation and interpretation. A registry that assigns purely on desire may preserve uniqueness while degrading the informational meaning of the record.

Distinct external policy and multi-homing are sensible operational proxies for genuine global need. They focus on circumstances in which an operator's choices cannot readily be represented inside a provider's system. The prospective clause avoids a catch-22 by allowing the number to precede the configuration it enables. Good-standing membership identifies the service counterparty, provides a means to keep contacts and fees current, and supports long-term record maintenance. Because planned topologies are varied, some human evaluation is unavoidable; no perfectly mechanical checkbox can anticipate every legitimate exceptional case.

RFC 1930 strengthens this defence by warning that exchanging external routing information or preparing for an indefinite future is not enough. A provider or private number can sometimes do the job. Requiring an applicant to explain why it cannot is not inherently an attempt to rule the applicant. It can be ordinary stewardship of a shared uniqueness record. The no-screening counterfactual reduces administrative discretion, but may produce gratuitous AS identities and lower-information routing state. That is a real cost, and a critique that ignores it is incomplete.

The 2006 amendment can therefore be defended as completion of an eligibility document rather than an expansion into regulation. On that account, the technical test already protected meaningful use, while the added line ensured that the recipient was an identifiable, current participant in the registry service. Staff inquiry allowed the registry to distinguish a concrete near-term need from a speculative request. The Board's ratification then authorised staff, within the private institution, to administer that combined test.

But this defence supports only an evidence-bounded ledger gate. It does not cure the missing conjunction. It does not explain what “reasonably short” meant, demonstrate equal treatment or supply reasons for any actual decision. Human evaluation is not a synonym for unbounded discretion. Its legitimacy within the service depends on tying each question to a published proposition, separating technical need from account status, recording reasons, permitting correction and preserving a bounded consistency review.

What the public chain does—and does not—show

The record is strongest at the policy level. It preserves the original post, named objections, AFRINIC's price and document clarifications, a meeting report of consensus, last-call dates, the Board ratification and instruction to staff, a contemporaneous comparative OR summary, and later consolidated wording. Together these materials establish a credible sequence from proposal to institutional adoption. They also establish the amendment's limited textual delta and expose the formal text's unresolved conjunction.

The record becomes thin precisely where policy became an individual service decision. There is no recovered 2006 request form, hostmaster checklist, evidence standard, applicant correspondence, decision template, acceptance or refusal ledger, reasons code, cure notice, elapsed-time target, escalation route, appeal path or post-ratification implementation bulletin. No named staff decision-maker or delegation is preserved. There is no documented applicant whose treatment can show whether policy and practice aligned.

These absences are an audit gap, not proof that AFRINIC kept no private records, acted in bad faith or harmed a named operator. Lost, confidential or unpublished material may have existed. The careful conclusion is limited to public verifiability: the surviving chain does not allow an outsider to reconstruct how staff resolved the ambiguity, which proof it accepted, why any request succeeded or failed, or whether similarly situated applicants received consistent treatment.

Several documents would materially change that assessment. A signed or minuted resolution with its full date, attendance, votes and attached text would strengthen the Board record. A final post-last-call instrument and implementation notice could establish the language staff received. A contemporaneous request form, hostmaster checklist and written conjunction guidance could define the burden. Redacted correspondence, reasons, correction records and review outcomes could show practice. Versioned aut-num records could connect assignment and maintenance actions without exposing confidential commercial detail.

Until such evidence emerges, later consolidated language must remain in its proper role. It can show that the clauses and an administrative record model endured. It cannot fill every 2006 blank. Likewise, modern analyses by BTW and NRS illuminate the boundary between registry coordination and sovereign enforcement; Heng Lu supplies the thin-coordination, local-verifiability and continuity frame; and LARUS identifies opaque registry decisions as an operator and infrastructure risk. These are first-class analytical contributions, but none is an eyewitness record of a 2005 list exchange, a Board vote or a 2006 application outcome.

A private ledger gate, not rule over the network

The institutional boundary is clearest when the decision is described from input to output. An organisation presented identity, service status and assertions about planned external routing. AFRINIC staff, acting under a Board-ratified private policy, could seek information and decide whether the case justified a globally unique ASN. A favourable result produced an assignment and registry record. An unfavourable result meant the registry declined to make that entry under the criteria before it. That is a meaningful service decision, but it is not a public adjudication.

AFRINIC's defensible role was that of a narrow Internet number-resource bookkeeper and uniqueness coordinator. It could protect deterministic uniqueness, maintain accurate records, assess whether an identifier was useful and administer its contractual relationship within applicable law. It was not Africa in institutional form. It was not a sovereign or legislature, not a government regulator or licensor, not a police force or prosecutor, not a court or punishment authority, and not a body empowered to confiscate or command the assets and conduct of nonparties.

The distinction does not belittle coordination. Accurate unique identifiers, current contacts and intelligible policy records can reduce costly confusion. A private institution may lawfully set conditions for its service and refuse a request that does not meet them. But usefulness and contractual capacity do not transform a database decision into an operating permission. Official statements about community, stewardship, recognition or authority prove what the institution said and did; they do not by themselves establish a public mandate.

The network's running reality remains outside the ledger. Operators configure routers, select commercial partners and decide their external policies. Remote networks decide, independently and voluntarily, which routing information they use, filter or disregard. A registry assignment does not compel reachability, and a registry refusal does not erase configurations already running. Institutional continuity and operator or protocol continuity are related but not identical. The ledger describes and coordinates a unique identity; it does not command the network into existence.

That is the line Resolution 200605.27 should bring into view. The Board made good standing an explicit condition of a private assignment service and told staff to implement it. It did not create the technical limbs, resolve their conjunction in the published formal text, or establish a public power over topology. The enduring governance task is to keep a useful necessity screen thin enough that evidence answers a uniqueness question rather than becoming an invitation to rule the applicant.