Summary

  • APNIC's 2026 survey data table says 20% of the 1,157 respondents in scope for Question 14 selected RDAP as a core registry service that they or their organisation use.
  • The question allowed several answers and combined personal with organisational use. It therefore does not supply a deduplicated count of organisations, a deployable-population denominator or an audited integration state.
  • A working service, aggregate query traffic, distinct source addresses, software availability and retirement of a WHOIS dependency are separate observations. None can silently substitute for the survey result.
  • Future reporting should present a small RDAP evidence stack that labels each numerator, denominator, observation unit and decision use.

One cell, several tempting stories

The most interesting number on page 17 of APNIC's 2026 survey appendix may be the smallest one in the service list. Whois was selected by 60% of respondents, the Internet Routing Registry by 58%, RPKI by 48%, reverse DNS by 37% and the Registration Data Access Protocol by 20%. Sixteen per cent selected none of them.

It is easy to turn the RDAP cell into a story of adoption: one in five has adopted the newer protocol. The source does not say that. Question 14 asked APNIC Members or Account Holders and people who belong to a National Internet Registry in the region which services “you or your organisation” use. It was a select-all question. One answer may describe a respondent's personal lookup habit, a production integration used by a whole company, a tool used occasionally by one team, or knowledge of an NIR service. The table cannot distinguish those states.

That limitation does not make the result weak. It makes the result specific. APNIC has learned that one fifth of the eligible respondent sample recognised enough current use to select RDAP. The same table reports 19% among APNIC Members and 26% among NIR Members, with subregional results ranging from 14% in Oceania to 27% in East Asia. Those are signals worth following. They are not deployment certificates.

The denominator is a respondent population

An adoption rate needs an observation unit. Is the unit a person, a member account, an organisation, a software installation, a configured client, an NIR service, an address lookup or a query? The survey uses respondents. Its wording permits the respondent to answer for either self or organisation. The published table does not expose an organisation-level deduplication rule, the number of organisations technically able to deploy RDAP, or the depth of any integration.

The select-all design matters too. The percentages are overlapping descriptions, not slices of a market that must add to 100. A respondent can use both Whois and RDAP, or use RDAP for registration data while retaining the IRR's RPSL functions. The 2021 APNIC Blog assessment made that coexistence explicit: RDAP was deployed across the RIRs and several NIRs, while Whois-formatted data and the IRR function still created legacy dependencies.

Question 15 offers priorities, but it changes the base. The ranking was asked of respondents who used the surveyed services, and the report gives a base of 890. Its RDAP ranking cannot be divided into the Question 14 population and presented as a conversion rate. It answers a different question about a conditional group.

Deployment, traffic and replacement are different layers

APNIC already publishes another kind of evidence. Its hourly Whois and RDAP files contain aggregate query totals, origin-ASN counts, query-type distributions and distinct source-IP counts. That can show workload composition at a given service. It still cannot count adopting organisations. One automated client can generate more traffic than thousands of occasional human users; one organisation can originate queries through many addresses or networks; shared infrastructure can represent many customers.

A working response from rdap.apnic.net proves a narrower fact: a particular query path returned a structured RDAP object at the observation time. RFC 9082 defines the query forms and RFC 9083 the JSON response structures. Protocol conformance and service reachability do not reveal how many member organisations have changed their workflows.

The February 2026 APNIC Blog review of the wider RDAP ecosystem used still other measures: server deployment, query volume, client use, available software and the withdrawal of WHOIS by some domain registries. Each can support an adoption argument when its scope is named. Combining them into one unlabelled percentage would destroy the useful differences.

Publish an adoption evidence stack

The next APNIC survey should retain the Q14 series; continuity has value. Beside it, APNIC could publish a compact evidence stack with four rows.

The first row would remain reported use: exact question, eligibility rule, respondent count, response mode, overlap rule and segment bases. The second would describe organisational integration through privacy-safe voluntary bands: lookup only, scripted client, operational dependency, customer-facing use or not known. The third would describe service observation: endpoint and NIR coverage, stable traffic measures, automation caveats and material classifier changes. The fourth would describe replacement readiness: which public-directory functions have moved, which data still originates in Whois workflows, and which IRR uses remain outside RDAP.

No row should be promoted as the master score. The EC may need the survey row when deciding whether to fund training. Engineering may need coverage and client evidence when planning support. An NIR may need integration detail. A decision about retiring a legacy interface needs residual-dependency and failure-mode evidence. The point is not to collect more data by default; it is to keep unlike evidence from borrowing authority from one another.

What 20% does and does not establish

The result does not establish that 80% of APNIC Members reject RDAP, do not know it, cannot reach it or prefer Whois. It does not measure service quality, protocol value, query share, regional endpoint coverage or migration readiness. It also does not contradict the existence of a functioning APNIC service or wider deployment by RIRs and NIRs.

It establishes a bounded and timely fact: under a particular 2026 survey question, 20% of the in-scope respondents selected RDAP. Treating that statement precisely is not statistical caution for its own sake. It lets APNIC decide what to measure next without turning one honest answer into five unsupported claims.

Primary sources