Summary

  • APNIC’s report totals registered IPv4 assignment instances, not individual addresses or active routes.
  • The monthly net increase of 660 hides much larger shifts between economy-code buckets, with no transaction log to explain why.

The newest file looks like a tidy verdict. APNIC’s 31 October snapshot shows 1,323,907 IPv4 assignment instances, 660 more than the file dated 1 October. IPv6 instances rose by 383 over the same comparison. A chart can turn those two numbers into a confident story about address growth before its author notices what the report actually counts.

The dates matter. The 1 October file covers records through 30 September; the 31 October file covers records through 30 October. Each is a daily summary. The comparison is between two snapshots of APNIC’s register, not a meter watching traffic move through the region.

APNIC’s assignment README says the report includes assignments in both its public Whois database and its private customer database. Its rows group records by an economy code, resource type and assignment size. For IPv4, value describes the number of hosts in a range; the range need not be a CIDR block. instances says how many assignments of that size appear in the group. The 660 is the net change in that last field, summed across IPv4 rows. It is not 660 addresses, 660 networks switched on, or 660 new customers.

That distinction is not pedantry. The same month’s buckets moved much more than the headline total. The ph code group gained 1,546 IPv4 assignment instances while au lost 956. The net across the whole file was still only +660. The file does not say whether those shifts came from new records, corrections, changed grouping or another administrative cause. It does not link them to a migration between countries or to any operator’s live network.

APNIC’s own documentation draws the hard boundary: the assignment report is not an authoritative account of assignments actively routed on the global Internet. The broader RIR statistics specification describes the files as snapshots without transaction or historical detail. APNIC also says its resource reports cannot accurately map an address to the country where it is currently being used. The code attached to an assignment is a register field, not a travelling geolocation beacon.

So what can the number do? It can describe how many assignment instances APNIC’s report records under its grouping rules on a dated snapshot. It can help a reader follow the register’s composition over time, provided the reader keeps the unit and the cut-off date in view. It cannot settle whether an address range is routed, reachable, carrying traffic, used by subscribers, or newly deployed. Those are different questions, and this file does not contain their answers.

The trap is that a small, clean number travels farther than its definition. “IPv4 assignments rose by 660 instances” is accurate. “The region added 660 addresses” is wrong by unit. “The Internet grew by 660 networks” is a second, larger leap. A neat table makes both leaps look respectable.

The missing evidence is specific: APNIC would need to publish a record-linked change history explaining why assignment buckets changed, and a separate dated routing-status source would be needed to connect registry records to active routes. Until then, the ledger can tell us what its rows say. It cannot tell us what the packets did.

Sources