Summary

  • APNIC's 3 November cumulative transfer original records 202.89.220.0–202.89.223.255 moving at 2026-11-02T23:38:39Z from Indonesia Network Information Center/dnk-id to Indonesia Network Information Center/SCBD-ID. The range is a 1,024-address /22. The preceding 3 November extended delegation file has a single 4,096-address parent; the 4 November file partitions it into a /21 and two /22s, with a separate opaque holder ID for the transferred /22.
  • APNIC's public WHOIS gives the exact transferred /22 its own SCBD-NET object, last modified at 23:38:47Z, eight seconds after the transfer-log timestamp. The same WHOIS answer also includes the older IDNIC DNK-NET /20. RDAP for that /22 returns only the IDNIC /20 object. The two lookup surfaces expose different levels of specificity for the same address.

This story begins with an oddly ordinary receipt. APNIC's dated transfer original gives the source as Indonesia Network Information Center/dnk-id, the recipient as Indonesia Network Information Center/SCBD-ID, and the exact interval as 202.89.220.0–202.89.223.255. Its timestamp is 2 November 2026 at 23:38:39 UTC. The slash names identify entries in the transfer record; they do not explain the transaction's commercial terms or the relationship between the two organizations. The interval itself is exact: 1,024 addresses, one quarter of the former /20.

The daily statistics show what the receipt leaves behind. The 3 November extended file contains one allocated Indonesian row beginning 202.89.208.0 with 4,096 addresses, the full 202.89.208.0/20. The 4 November file replaces that row with 202.89.208.0/21 holding 2,048 addresses and two 1,024-address /22s beginning 202.89.216.0 and 202.89.220.0. They are disjoint and exactly refill the parent. The first two share opaque ID A92AB1BD; the transferred /22 bears A92B4C5C within the same 4 November file. APNIC's extended-file README says one opaque ID identifies one holder within a file, while warning that the code need not remain constant between versions. The evidence for distinct holders is the split inside the 4 November file, not a comparison of code strings across days.

The old 20050419 delegation date survives all three new rows. APNIC's statistics specification says the daily file is a current-state snapshot with no transaction history. Its old date is the resource's allocation date, not a denial of the 2026 transfer. The transfer log supplies the event clock; the daily file supplies the partition. These two official sources agree on the transferred quarter without pretending that a 2005 date describes 2026 activity.

WHOIS adds names and a narrower cut. Search the three addresses in APNIC's public interface and the first inetnum objects are the exact retained /21, retained /22, and transferred /22. The first two are DNK-NET; the last is SCBD-NET. All three say last-modified: 2026-11-02T23:38:47Z, eight seconds after the transfer timestamp. Those are two recorded administrative times, not a measurement of global propagation. Each WHOIS answer then includes the older, encompassing DNK-NET /20 marked source: IDNIC, last modified in 2011. The current child and the inherited parent are both visible, if one reads past the first record.

RDAP is where the neat partition disappears. The transferred /22 query, the middle /22 query, and the left /21 query each return the same 202.89.208.0–202.89.223.255 DNK-NET parent. Their notices say the object came from IDNIC and was partially mirrored by APNIC. The exact APNIC-source child objects visible in WHOIS are absent from these RDAP responses. That is a query result, not a finding about who actually operates a router or who will answer an abuse message. A new exact-child RDAP response would change the result immediately.

APNIC's transfer-log format expressly suggests testing a recipient's holding by looking up the prefix in the Whois Database. Here that test works when the operator can read the WHOIS response and distinguish its APNIC child from its IDNIC parent. A reader using only the RDAP response gets the parent and its older label. The registry has published enough facts to reconstruct the transfer, but it has made the reconstruction depend on which public door the reader opens.