Summary
- The IPv4 address registry run under IANA's assignments process defines two different relationships: space delegated entirely to a regional registry, and legacy space allocated before regional registries existed and now administered by one, including maintenance of WHOIS and reverse DNS.
- Two African /8 blocks carry May 1993 dates with AFRINIC named as the administering registry, while 041/8 carries April 2005 as an allocation to AFRINIC itself — a twelve-year gap between the date on the space and the date on the institution's own delegation.
- Custody is what the handover settled. The records that would show what else moved — a bilateral instrument, migration mechanics, transfer records, certification continuity — were not obtained in this review, and that gap is a governance fact rather than a research footnote.
There is a sentence in the address registry that reads like housekeeping and behaves like constitutional law. It appears under a legend called LEGACY: the space was "allocated by the central Internet Registry (IR) prior to the Regional Internet Registries (RIRs)", and it "is now administered by individual RIRs as noted, including maintenance of WHOIS Directory and reverse DNS records". A few lines away, under ALLOCATED, the registry says something shorter and harder: "delegated entirely to specific RIR as indicated" (IANA IPv4 address space registry).
Those two sentences, read together, do something no communiqué does. They separate the question of who gave out an address from the question of who keeps the record of it. And once you hold that separation in your hand, the 2004–2005 transfer of African registry service stops looking like a single event with a completion date and starts looking like two events with two different clocks: one that finished, and one that is still being administered.
The evidence for that is not a leaked annex or a disputed minute. It is a date column.
Two dates that do not agree
In the same registry file, two African blocks are listed with a 1993 date and the words "Administered by AFRINIC": 154/8 and 196/8, each carrying 1993-05. A third block, 041/8, is listed as ALLOCATED to AFRINIC with the date 2005-04. Two more blocks follow later in the file as outright allocations to AFRINIC — 105/8 in 2010-11 and 102/8 in 2011-02 — but they are not the interesting ones, because by then the institution was the one doing the allocating and its name on the line is unremarkable.
The interesting pair is the first: space dated May 1993, administered by a registry whose own recognition as the African regional registry is dated 2005 by the outgoing registry's announcement. Read the archive in the order it was written and you get a tidy sequence — an institution arrives, the region's records change hands, service continues. Read the date column instead and you get something stranger and more useful: a registry recognised in 2005 holds the maintenance obligation for address decisions taken twelve years before it had any regional standing. The 1993 date is not a typo in the file.
It is the whole point of the LEGACY category.
What "administered by" actually obliges
The registry does not leave the word vague. Legacy space, it says, is "now administered by individual RIRs as noted, including maintenance of WHOIS Directory and reverse DNS records", and assignments from those blocks "are distributed globally on a regional basis" (IANA IPv4 address space registry). That is a duty list, not a courtesy. WHOIS and reverse DNS are the two records that tell an operator, a court, a counterparty or a routing engineer who stands behind a prefix. If the administering registry lets either decay, the address still works and the accountability around it stops working.
Now put that beside what the registry says about allocation: delegated entirely. A delegation makes the registry the source of the decision — the body that said yes, on its own policy, on its own terms, on its own date. Administration does none of that. Administration makes the registry the keeper of someone else's decision. The space is not re-issued, the holder is not re-founded, and the original date is not quietly revised to match the new custodian's birthday.
That is a distinction with teeth, and it cuts against the way registry history is usually told. A transfer narrative implies that authority moved wholesale: the region's addresses went from one registry to another, so the new registry now stands where the old one stood. The registry's own legend says something narrower. What moved was the maintenance. What did not move was the allocation date, the original decision, or the standing of the organisations holding space under it.
The arithmetic is not the argument, but it is not nothing
A single /8 contains 16,777,216 IPv4 addresses. The two 1993-dated blocks together therefore cover 33,554,432 addresses, and the registry records one allocation date for each: May 1993. The first block the file records as delegated to AFRINIC itself, 041/8, covers 16,777,216 addresses and carries April 2005.
So the space AFRINIC administers on 1993 paper is twice the size of the first block the registry says was delegated to AFRINIC. Anyone tempted to treat legacy space as a rounding error in the region's inventory should sit with that ratio for a moment. The older, inherited, never-allocated-by-this-registry category is not the tail of the ledger. On the evidence of this file, it is the larger half of the institutional story.
There is a discipline point here too, because the temptation with numbers like these is to inflate them into a total. The registry file offers five rows with dates and designations, not a total for space transferred in 2004–2005, and this article does not invent one. What those rows establish is the shape of the thing: inherited space with an old date, delegated space with a new one, and a maintenance obligation attached to the former that is named in the same document that names the designations.
Recognition is not transfer, and the record keeps them apart
In the recognition trail there are two ICANN records dated 31 March 2005: a board-resolutions page and a press release (ICANN board resolutions, 31 March 2005; ICANN press release, 31 March 2005). Their operative wording was not obtained in this review, so nothing here turns on how those instruments phrase the decision. Their existence is enough to fix a date and to make a structural point that the rest of the record keeps making: recognising a registry and transferring a registry's work are different instruments answering different questions.
One creates standing among registries. The other moves the maintenance of records. Conflating them produces the sentence that has done the most damage to plain understanding of what happened in 2005 — that the transfer was completed, therefore everything about the region's older addresses was settled at the same moment. The registry's own date column refuses that sentence. If everything had been settled in 2005, the 1993 dates would have been rewritten, or the blocks would have appeared as fresh allocations. They appear as neither. They appear as legacy space under administration, which is precisely the residue.
The outgoing registry's own account supplies the transition narrative. RIPE NCC announced that AFRINIC was recognised as the regional registry for Africa in 2005 and that the transition involved transferring management of IP address space for the African region to AFRINIC (RIPE NCC: AFRINIC starts operations). In a later account it described itself as having been the de facto registry for Africa before that recognition, and noted that legacy registrations remained (RIPE NCC: AFRINIC becomes the fifth regional internet registry).
Read those two statements carefully, because they are careful statements. "Management" is the word used, not ownership, not title, not re-allocation. And "legacy registrations remained" is not a loose phrase about old paperwork; it is the same phenomenon the registry file records in its date column. The outgoing registry is describing a handover of operational custody and saying, in the same breath, that some registrations stayed where they were in every sense that matters to their holders.
The receiving registry's side of the desk
There is a page at AFRINIC addressed to legacy resource holders (AFRINIC: Legacy Internet Number Resources). Its operative rules were not obtained in this review, so this article cannot tell you what AFRINIC currently requires of a legacy holder — whether documentation demands differ, whether transfer requests are treated differently, whether certification is offered on the same terms. That is a real gap and it is stated here rather than papered over.
But the existence of the page is itself informative, in a modest and defensible way. A registry does not maintain a dedicated channel for legacy holders unless legacy holding is a distinct administrative category on its side of the desk. The category exists because the registry's own inheritance created it: space it administers and did not allocate, held by organisations whose relationship to the registry began as an obligation rather than as a membership application.
That is the institutional asymmetry at the centre of this story. For space it allocated, a registry is the author of the relationship. For space it inherited, it is the executor. The same staff, the same database, the same policy process, two entirely different sources of authority — and a holder who cannot tell which one applies to them cannot tell what they are entitled to ask for.
What the record does not say
The boundary of this review matters more than usual here, so it is worth setting out plainly. Five of the six records cited in this article were identified but their bodies were not obtained in this review. That includes the bilateral transition instrument, if one exists in public form: no agreement, annex or schedule between the two registries was obtained. Nothing here establishes whether African registration objects were migrated, copied or recreated during the handover. Nothing here establishes how address-transfer records were treated, during the handover or after it.
Nothing here establishes whether resource certification and reverse DNS delegation for the 1993-dated blocks continued without interruption. No count of legacy registrations held by African organisations was obtained. No total figure for IPv4 space moved in 2004–2005 was obtained.
Those are not rhetorical hedges. They are the specific items that would convert this argument from a reading of the registry's categories into an audit of the handover, and they are exactly the items a governance-minded reader should demand from whoever claims the handover was fully accounted for.
What the record does say, and says in the registry's own voice, is that a maintenance obligation attaches to administered legacy space, and that this obligation names WHOIS and reverse DNS. That sentence is the load-bearing wall. Everything else in the handover — the communiqués, the recognition, the transition period, the service agreements — sits on top of it.
Why the residue binds the present
A registry's authority over space it did not allocate is real but bounded, and the boundary runs through records rather than through addresses. Packets do not care which registry's name is on the delegation. Counterparties do. Due diligence does. A transfer request, a dispute over who may authorise a routing change, a certification claim that a prefix is held by the organisation asserting it — each of those turns on a record that some registry maintains and someone else relies upon.
That is why the residue is not a historical curiosity. The 33,554,432 addresses in the two 1993-dated blocks are not dormant inventory; they sit inside the region's operational estate, under a registry that inherited the duty to keep their records legible without inheriting the original decision that created them. When the maintenance is clean, nobody notices. When it is contested, the contest is about documents, and the documents belong to two institutions with different histories.
And there is a further consequence that the recognition-versus-transfer distinction makes visible. If the handover settled custody and left the documentary treatment of transfers and certification open, then arguments about AFRINIC's present registry conduct cannot be resolved by pointing at 2005. The recognition date answers a question about standing. It says nothing about whether a given legacy record is complete, whether a given transfer was recorded, or whether a given certification claim is continuous. Those questions are answered by records, or they are not answered at all.
The shape of the finding
It would be satisfying to close with a resolution: the handover was completed, or the handover failed, and the region's registry history can be filed accordingly. The registry's own text does not support either ending.
What it supports is narrower and more durable. Recognition and transfer are separate instruments. Allocation and administration are separate relationships. The handover completed the second of each pair and left the first intact, which is why a 1993 date still sits under a 2005 institution and why that arrangement is not an anomaly to be corrected but the actual design. A receiving registry inherits the duty to keep another institution's older decision legible. It does not inherit the power to re-issue what that decision covered, and it does not re-found the standing of the holder.
There is one more thing worth naming, carefully, because it is a feature of the record rather than an accusation against anyone. In the material this review could reach, the handover is documented largely by the outgoing registry's account of its own departure — its announcement, its retrospective — while the receiving registry's own operative rules for legacy holders were not obtained, and the deciding body's operative wording was not obtained. A transition described mainly by the party leaving is a transition whose remaining questions are documentary by construction.
That is the residue. Not a missing spreadsheet, not a conspiracy of silence: a category of space whose dates precede the institution that maintains it, a duty list that names WHOIS and reverse DNS, and a set of documents about transfers and certification that a reader should keep asking for. The addresses were never re-issued. The obligation to remember them was.
The entity record for this transition, with its directory context, sits at RIPE NCC to AFRINIC transition in the BTW directory Only for Strategic Circle.
Member Briefing
Deeper Profile Context
Sign in with the right membership level to unlock the full briefing and source notes.
Only for Strategic Circle
Strategic Circle
Open to all readers. Unlock profile briefings after joining and signing in.
Join Strategic CircleOnly for Leadership Alliance
Leadership Alliance
For qualified IP-asset owners and management; sign in to unlock alliance briefings.
Join Leadership Alliance





