Summary

  • Africa's regional internet registry was recognised by ICANN in April 2005, and that recognition is described as following completion of a transition process rather than as the transfer itself.
  • The outgoing registry, RIPE NCC, says it had provided African-region services including address allocation, registry database management and reverse DNS delegation, then facilitated migration of those responsibilities and records — a separate act, by a different institution, on different authority.

Two institutions appear in the name of this subject, and only one of them was in a position to declare anything. ICANN recognised AFRINIC as the fifth Regional Internet Registry in April 2005, an act whose public trace is an announcement framed around the completion of a transition process. RIPE NCC, which had been running registry services for the African region, describes the other half: it provided address allocation, registry database management and reverse DNS delegations, and upon AFRINIC's recognition it facilitated the migration of those responsibilities and records to the new registry (ICANN; RIPE NCC).

That pairing looks unremarkable until you try to audit it. Recognition and migration are not two descriptions of one event. They are two acts, performed by two institutions, on two different kinds of authority, and they generate two different kinds of evidence. The first is a designation: a body with regional standing says that another body is the registry for a region. The second is custody: an organisation that held registration records, a database and delegation responsibilities stops holding them and another organisation starts.

Designation is the kind of act that produces a public artefact almost automatically, because its whole purpose is to be legible to everyone outside the room. Custody is the kind of act that produces operational artefacts — record populations, database migrations, delegation changes, correspondence with the holders whose entries move — none of which is obliged to be a public document. Doctrine on this subject draws the line sharply here: the obtainable record may be described as uneven across the two acts, but uneven documentation is not unequal integrity, and it is not evidence that custody moved without accountability.

So the honest question is narrower than the headline version, and more useful. When a registry succession is certified by one body and executed by another, what would an outsider need in order to check that the second act happened the way the first act implies?

What recognition actually certifies

The ICANN announcement is not a transfer instrument. It is a statement that a transition had been completed and that the body which emerged from it was the regional registry. Read that way, the announcement is a certification of someone else's work. ICANN did not allocate African addresses, maintain the registry database or hold reverse DNS delegations; on RIPE NCC's own account, RIPE NCC did those things. The recogniser attested to the completion of an operational process it did not perform.

That is not a criticism of ICANN. Certification by an external body is exactly what a delegation of regional authority should look like, and a recogniser that tried to perform the migration itself would be doing something else entirely. But it does mean the recognition document cannot carry the evidentiary weight that its prominence invites. It tells you that the transition was judged complete. It does not tell you, in any detail, what completed means.

Three named functions, and what they imply

RIPE NCC's account names three classes of function: address allocation, registry database management and reverse DNS delegation. That enumeration is the most concrete thing in the obtainable record, and it is worth taking seriously as an enumeration rather than as a certified inventory.

Each of the three has a different failure mode, which is why moving them together is harder than moving them in sequence. Address allocation is prospective: from a cutover, the receiving registry decides who gets space. Registry database management is cumulative: it concerns entries that already exist, created under the outgoing registry's procedures, with contact records and status values that may not map cleanly onto the receiving registry's schema.

Reverse DNS delegation is outward-facing: it is visible in the DNS itself, to anyone who queries it, and it is the one part of the handover that a third party can partially observe without being given access to anything.

A handover that names all three but does not date them individually has a specific blind spot. Allocation could have stopped on one date, database migration could have run over months, and delegation changes could have continued afterwards. Without a cutover date, a reader cannot tell whether a registration created in the second half of 2005 was made under the outgoing registry's rules or the incoming one's. That matters, because the rights and obligations attached to a registration follow the rules in force when it was made.

Why the second act is structurally quieter

There is a temptation to read the asymmetry as a choice: one institution published, the other did not. The structural explanation is more plausible and less interesting, which is usually a sign it is right. A designation is addressed to the world; it is supposed to be found. A migration is addressed to a working relationship between two organisations and to the holders whose entries move; it is supposed to work. The first is a speech act. The second is a project.

Projects leave internal records — migration plans, reconciliation reports, exception lists, correspondence about the entries that did not move cleanly. Those are exactly the documents that would let an outsider verify the succession, and they are exactly the documents that tend not to be published, because publishing them was never the point. This is not unique to internet registries, and it is not a scandal. It is the ordinary gap between the act that confers standing and the act that transfers custody.

The consequence, though, is real. If the supporting instruments stay outside the public record, the practical result is reduced independent auditability. Anyone who wants to know how African number-resource custody actually changed hands must rely on institutional accounts rather than on independently available instruments. That is a statement about what the record can support. It is not a statement that accountability was absent, and the difference between those two sentences is the whole argument.

What is not established here

The handover instrument or agreement between RIPE NCC and AFRINIC was not obtained in this review. The text of the ICANN Board resolution of April 2005 was not obtained in this review. IANA address-space reassignment entries that would show individual block movements were not obtained in this review, and no cutover date for African-region registry service was obtained in this review. Repeated retrieval attempts for these instrument classes ended in provider permission, capacity and quota failures, which is a fact about this research process, not a fact about the instruments.

Nothing above should be read as a finding that no agreement exists, that RIPE NCC published nothing about the transfer, that recognition and operational handover occurred on the same date, or that any particular address block was or was not re-registered. The specific day of the April 2005 recognition is itself reported inconsistently across indexed renderings, so only the month is bound here.

Why the distinction matters beyond 2005

Registry successions are rare enough that each one becomes a precedent for the next. The pattern visible here — external designation documented in an announcement, operational custody documented in institutional retrospect — is the pattern any future succession will probably follow, because it follows from what each act is for.

For holders of number resources, the practical question is not who was recognised. It is whose rules applied to their entry on the day it was created or changed, and whether that entry survived the migration intact. For a receiving registry, the question is whether it inherited a reconciled population or an unreconciled one, because legacy entries become the disputes of the following decade. For the global system, the question is whether a recognition can be verified after the fact by anyone who was not in the room.

None of those questions is answered by an announcement. All of them are answerable in principle, by instruments that a registry succession could publish as a matter of course: a dated cutover, a statement of the record population transferred, a description of the treatment of entries that did not transfer cleanly, and a pointer to where delegation changes were recorded. That is not a demand for transparency as an abstract virtue. It is a description of the minimum evidence that would let a successor recognise itself in its predecessor's work.

Until that evidence is routinely published, the honest description of a registry succession is the one this subject's own record supports: the receiving registry publishes its institutional history (AFRINIC), the recognising body publishes the designation, the outgoing registry publishes a retrospective account of what it used to do, and the custody transfer itself must be inferred from the shape of the three.

That inference is worth making carefully, and it is worth naming as an inference. The alternative — treating the announcement as proof that the handover was verified — mistakes the legibility of the first act for the completeness of the second.

For the full institutional record on this subject, see the BTW directory entry Only for Strategic Circle.