Summary

  • IANA records two adjacent 1,024-number blocks for AFRINIC: AS328704–AS329727, registered in 2020, and AS329728–AS330751, registered in February 2026.
  • AFRINIC's dated statistics record AS329727 as allocated on 16 April and the first three numbers of the new block as allocated on 17 April. The daily publication files reveal those changes one file later.
  • In the 10 September file, only 972 values from the older block appear as allocated; 52 are absent. The new block contains 81 such records. Reaching the end of a numeric range therefore does not prove that every earlier value was delegated.
  • Nothing in the reviewed record explains the 52 absences. A public block-reconciliation receipt should count and classify inventory states without exposing private applications or confusing registry state with routing and use.

Two consecutive numbers, two different institutional acts

The line between 329727 and 329728 is arithmetically trivial. Add one. Institutionally, it crosses a boundary.

IANA's Autonomous System Number registry places AS328704 through AS329727 in a block assigned to AFRINIC on 5 March 2020. It places the immediately adjacent range, AS329728 through AS330751, in another 1,024-number block assigned to AFRINIC on 25 February 2026. AFRINIC announced the newer allocation on 27 March, saying that the range had been added to its inventory.

Those are upstream records. They describe a pool moving into the regional registry's custody. They do not say that all 1,024 numbers had already been assigned to networks. IANA makes the distinction explicit: it allocates AS-number blocks to the RIRs, and the RIRs then allocate or assign AS numbers to network operators under regional policy.

AFRINIC's delegated-statistics files describe that downstream layer. The exchange specification defines an ASN row as a starting number and a count. It defines the date as the date on which the producing RIR made the allocation or assignment. It also warns that the status is the act of the RIR producing the file, not a sub-assignment made by another agency. That modest definition is the key to reading the boundary correctly.

The 10 September row for ASN 329727 identifies Nigeria as the country where the holder is legally based, gives 16 April 2026 as the allocation date and marks the number allocated. The next three rows begin the new IANA block: 329728 with Ghana, 329729 with Kenya and 329730 with Tanzania, all dated 17 April and all marked allocated.

It is a remarkably clean visual sequence. It is not a complete inventory statement.

The daily files show when the public record changed

Three archived files make the handoff observable without requiring a guess about website caches.

The file whose header date is 16 April does not contain ASN 329727 or ASN 329728. The 17 April file contains 329727 and retains 16 April in that row's allocation-date field; it does not contain 329728. The 18 April file contains both sides of the boundary and adds 329728, 329729 and 329730 with allocation dates of 17 April.

This one-day offset is not a contradiction. A row may describe an allocation made on one date and become visible in the next daily publication. The distinction matters because “date in the record” and “date on which this particular public file first displayed the record” answer different questions. One describes the registry act; the other describes public observability.

The transition can therefore be stated carefully. AFRINIC records the last number of the older block as allocated on 16 April. It records the first numbers of the newer block as allocated on 17 April. The public daily files expose those states on 17 and 18 April respectively. There is no need to manufacture a more precise hour, transaction order or service-level conclusion.

The count defeats the exhaustion shortcut

Now count the older block rather than looking only at its highest visible number.

AS328704 through AS329727 contains exactly 1,024 integer values. In the 10 September delegated file, 972 of those values appear in one-number rows marked allocated. Fifty-two do not appear. The visible allocation share is 94.921875 per cent; the absent share is 5.078125 per cent.

The absences are not a clean unused tail. They are scattered through the range. The file contains values above many of the gaps. Its last absent value is ASN 329708, while every number from 329709 through 329727 appears. A chart that reports only the maximum visible ASN would therefore place the pointer at the end of the block and erase 52 unresolved states behind it.

The newer block offers the complementary view. By 10 September, the file contains 81 one-number allocated rows between AS329728 and AS330751. It does not contain 943 values in that range. The first present value is AS329728, dated 17 April; the last is AS329808, dated 8 September. Those dates are 144 days apart. Again, this describes the file, not a demand curve. It does not show how many applications arrived, how long they waited, why a particular number was selected, whether a number is routed, or whether an assigned network is active.

The arithmetic is useful precisely because it is narrow. It disproves a claim the record cannot support: that crossing the highest number in the older block means all earlier numbers had been delegated. It does not replace that false certainty with a new story about the 52 absences.

“Absent” is a finding, not a status

There is a dangerous temptation in registry analysis to turn missing data into a named administrative state. If an ASN does not appear in a delegated file, one might call it free. Or reserved. Or returned. Or withheld. Or broken.

None of those labels follows from this evidence.

The exchange format tells us what a present allocated row means at the RIR layer. It does not provide a complete public vocabulary for every internal inventory condition. The source set reviewed here does not join each absent number to an application, reservation, hold, reclamation, fraud review, correction or policy exception. AFRINIC's inventory page lists the blocks it manages, not the disposition of every value inside them.

“Absent from the 10 September delegated file” is therefore the strongest defensible description of the 52. It is testable, reversible and bounded to a named source. If a later file adds one of those numbers, the observation can change without rewriting history. If AFRINIC publishes another state, the analysis can incorporate it without pretending that the state was already known.

The same discipline applies to country codes. The specification says cc identifies the country where the resource holder is legally based, while warning that it cannot reliably locate where the resource is used. Ghana, Kenya, Tanzania and Nigeria in the boundary rows are registration facts, not maps of routers, traffic or customers.

Why the sequence still matters

Rejecting the exhaustion shortcut does not make the boundary meaningless. The clean 16-to-17 April succession shows that AFRINIC could begin drawing from the newer IANA block while the public downstream file still left unresolved values in the older one. That is a normal possibility in many inventory systems. Sequential identifiers are not necessarily a stack that must be emptied in perfect order.

But a public institution should not let a convenient head number stand in for its inventory story. Operators, policy participants and other registries may use the highest visible ASN as a quick signal of activity. Without a denominator, the signal silently combines at least three different questions: how far the sequence has advanced, how many numbers are publicly recorded as delegated, and how many remain in each administrative state.

Those questions affect different decisions. Sequence advancement helps identify which IANA block is now being used. A delegated count helps reconcile public records. An inventory-state count helps test whether more upstream space is needed. Routing visibility helps study operational appearance. Treating one as a proxy for all four creates confidence without evidence.

This is not an allegation that AFRINIC made an improper request or concealed usable numbers. The IANA eligibility process and the timing of the 2026 upstream allocation are separate subjects. Nor is it evidence that any member was denied an ASN. The narrow governance problem is that the public record makes the sequence easy to see and the denominator possible to calculate, but does not explain the administrative status of what the calculation leaves behind.

A block-reconciliation receipt

The remedy can remain compact. AFRINIC need not publish applications, commercial plans, security reviews or member correspondence. It can publish a versioned receipt for each IANA ASN block.

The receipt would name the upstream range, size and IANA registration date; identify the dated delegated file and its cryptographic hash; state how many numbers are present in allocated or assigned rows; state how many are absent; and publish either the absent set or a stable digest and reason-code totals. It would distinguish the allocation date inside a row from the first daily file that exposed the row. It would define every public state it uses and give the effective date of corrections.

Most importantly, it would include two negative declarations. First, a downstream registry row is not proof that the ASN is visible in BGP or used by an active network. Second, absence from the delegated file is not by itself proof that the number is available.

With that receipt, the 2026 boundary becomes easy to audit without becoming dramatic. Readers can see 1,024 upstream numbers, 972 older-block downstream rows, 52 unresolved values, then 81 new-block rows. They can reproduce the count, identify what changed between daily files and stop at the edge of the evidence.

Sources