Summary

  • In ARIN’s 1 November transfer-log snapshot, 14 ARIN-to-other-RIR resource rows dated 21–29 October contain 77,824 IPv4 addresses, equal to 304 /24s, plus one ASN-only row.
  • ARIN says request totals need not match transferred-block counts. Its downloadable rows have no public request key, so the count of confirmed transfer requests cannot be reconstructed from this slice.

The easiest way to exaggerate an address market is to count the wrong thing with great confidence.

ARIN’s 1 November 2026 transfer-log snapshot contains 14 records with transfer dates from 20 October onward, ARIN as source RIR and another RIR as recipient. Six rows point to APNIC and eight to RIPE NCC. Add the IPv4 ranges in the transfer_set fields and the result is 77,824 addresses: 24,576 toward APNIC and 53,248 toward RIPE NCC. In /24 units, those are 96 and 208, or 304 together.

Fourteen is a count of resource rows. Three rows dated 21 October name the same Globenet Cabos Submarinos Americas source and Bharti Airtel recipient at the same instant. Two more rows share a St. Catherine University-to-Broadband Ltd timestamp. Grouping the 14 by timestamp, source, recipient and destination RIR produces eleven clusters. That grouping is useful for inspection; it is not an official request count. The JSON provides no public transfer-request identifier that would prove whether same-time rows belong to one case.

ARIN’s own specified-transfers statistics put the warning in plain sight: totals for 8.3 and 8.4 requests are not expected to equal the number of individual blocks transferred, because requests often include multiple blocks. The registry has already told readers that the denominators differ. The downloadable NRO-format log preserves resource records, dates and parties; it does not offer a public bridge from every resource row to its request.

The row most likely to vanish from an IPv4-only spreadsheet is also the clearest reason to separate resource types. On 28 October, ARIN records ASN 22665 moving from Ungage Capital LLC to Swiftlet Labs Ltd in the RIPE NCC region. That row has no IPv4 transfer_set. ARIN’s inter-RIR guidance expressly covers both unused IPv4 space and ASNs under section 8.4. A zero in the IPv4 column is therefore not a failed transfer; here it is a different number resource.

The other records have concrete scale without telling us their price. St Mary’s University-to-SalesForce.com accounts for 2,048 addresses toward APNIC on 21 October. Globenet-to-Infomaniak Network SA accounts for 16,384 toward RIPE NCC on 28 October. Leaseweb USA-to-Leaseweb Deutschland GmbH accounts for 2,048 toward RIPE NCC on 29 October. The log records the resource, named parties, registry direction and transfer date. It does not record the consideration, the reason for each move, the current route or how a recipient will use the space.

That is why the unit matters. An address total measures the amount of IPv4 represented in the listed ranges. A row count measures the records in one registry’s JSON. A request count measures cases handled by the registry. An ASN count measures another class of resource entirely. Put any one of these above a chart marked “deals,” and the chart acquires a certainty the source never supplied.

The straightforward fix is a privacy-safe grouping key in the public log, plus explicit resource-type totals. It could identify which resource rows belong to the same completed request without exposing the private application file or commercial terms. ARIN already publishes both the warning and the records. Joining them would let readers count what they say they are counting.