Summary

  • APNIC's 2025 annual report set a target of eight M-Root server deployments and reported four. Its New Zealand before-and-after measurement fell from 51.7 to 13.6 milliseconds; the Sri Lanka measurement is marked unavailable.
  • APNIC helped identify sites and provide administrative support while WIDE Project and JPRS operated and deployed M-Root. APNIC's stated 2026 role narrows to site introductions, with administrative support moving to JPRS. The missing Sri Lanka receipt crosses that handoff.

The sharpest line in APNIC's 2025 annual report is barely a line at all: “M-Root Sri Lanka measurement unavailable.” On the same table, New Zealand has a before figure of 51.7 milliseconds and an after figure of 13.6. A reader can see a reported improvement for one deployment and no comparable observation for the other. APNIC also says it provided administrative support for deployments in both countries. Two named places, one measured result, one blank cell.

The table was supposed to record a larger programme. Its target was eight root servers deployed in 2025 and six new sites identified. The actual column reports four servers deployed, four sites identified in the Asia Pacific and three outside it. Those are different counters: a site introduced is not a server installed. Calling the seven identified sites a substitute for four missing deployments would turn a pipeline into a delivery receipt. APNIC's own columns keep them apart.

Why insist on the distinction? M-Root is a DNS root service, not a promotional pin on a map. A new instance has to be hosted, connected, announced and observed. APNIC's account of the 2020 collaboration names WIDE Project as the root-server operator and JPRS as a joint operator and deployer. APNIC's part was to help expand the footprint and find hosts. The three organizations could each truthfully report activity while a site still waited for its first useful public measurement.

New Zealand at least has a number in the annual report: 51.7 to 13.6 milliseconds as measured with RIPE Atlas. The table does not give the probe set, observation window or distribution, so its improvement belongs to that reported measurement, not every user in the country. Sri Lanka has no equivalent number in the table. One more sentence would change the story: a dated, comparable pre-and-post measurement for that site. APNIC did not publish it in this scorecard.

The timing makes the blank more consequential. In September 2025, APNIC described a narrower Development role: it would identify and introduce potential M-Root sites, while administrative support moved to JPRS with APNIC Foundation assistance. Its annual report repeats the transition. A site-introduction team can hand over a host contact. It cannot, by that act alone, show that root-service response improved. The receiving administrator and the operators need to make the deployment record and measurement meet.

APNIC's January 2026 engagement recap lists new M-Root and measurement infrastructure around Sri Lanka and New Zealand. That wider account says something happened on the ground. It does not fill the annual report's Sri Lanka latency cell or reconcile the four delivered servers with the eight-server plan. The useful question is narrower than “Did APNIC do good work?”: which sites were installed, on what dates, and which observations belong to each one?

There is a respectable outcome in the New Zealand line. A reported fall from 51.7 to 13.6 milliseconds gives an operator a measurement to interrogate. The Sri Lanka line gives an operator a task. The four-of-eight line gives the programme a schedule question. As APNIC steps back from administration, those three items should travel together in the handoff folder. Otherwise the easiest thing to pass on is the announcement, and the hardest thing to recover is the proof.

Sources: APNIC 2025 annual report; APNIC's September 2025 Development transition; APNIC's M-Root collaboration account; APNIC's January 2026 engagement recap.