APNIC reports two China ROA coverage figures: above 80% and 68%

Summary

APNIC’s September 2026 EC minutes say China’s ROA coverage rose from 2% to over 80% in the first half of the year; the attached Q2 report says 2% to 68% as at 30 June.
APNIC Labs separates route-count coverage from address-span coverage. The minutes never identify which measure sits behind the higher figure.

The APNIC Executive Council minutes for 6–7 September carry two official figures for China’s ROA coverage. In the Director General’s update, APNIC says the rate climbed from 2% to over 80% during the first half of 2026. The same PDF attaches the Q2 Activity Report. Its H1 RPKI section records a rise from 2% to 68%, with all figures dated 30 June.

That is a gap of more than 12 percentage points inside one public packet. The Afghanistan comparison is identical in both places: 73% to 94%. China’s starting value also matches. The packet gives no date or denominator beside the higher China endpoint, so readers cannot tell whether APNIC is comparing two measures or whether one figure needs correction.

The missing label matters because APNIC Labs publishes two different kinds of ROA coverage. Its country table defines “Route Objects” as the share of advertised routes covered by a ROA and “Addresses” as the share of advertised address space covered by one. Route counts and address space are different denominators. The minutes do not say which series produced the figure above 80%, or whether it comes from a snapshot later than 30 June.

A ROA records which autonomous system is authorized to originate a prefix. APNIC’s RPKI guide describes Route Origin Validation as the separate act of checking BGP announcements against those ROAs. Operators then decide how routers handle each result. APNIC’s Q2 report says concerns about affecting customer routing still hamper ROV deployment by large operators. A rise in ROA coverage therefore belongs to the signing layer; it is not a count of networks filtering invalid routes.

That distinction sits inside a real deployment programme. APNIC’s Q2 report groups training, technical assistance, member service, promotion and meetings under its 2026 IPv6–RPKI effort. A percentage used to summarize that work needs the same care as the work itself: the date, denominator and counted route state should travel with the result. Otherwise a higher number sounds like a stronger outcome while leaving its unit off the page.

APNIC can settle the question with one public row: the measure behind “over 80%,” its snapshot date, and the numerator and denominator. If the two figures cover route count and address span, labeling both will make the difference useful. If they cover the same series and date, the record needs one corrected value. Until then, the minutes show progress in both accounts but do not provide a comparable China time series.