The bundled WHOIS table

Most of whois.json is generated rather than hand-maintained, by scripts/refresh-whois.py. It walks IANA’s list of delegated TLDs, asks whois.iana.org which server serves each one, and then — the part that matters — tests every answer before believing it. For each TLD it asks the server about a sixteen-character random label nobody can have registered, and about a name the DNS proves is registered because it has NS records. A server only reaches the table if the first reads as AVAILABLE and the second reads as TAKEN, judged by a port of ds’s own classifier.

The available needle is then chosen from a list of registry phrasings vetted by hand, and kept only if it is absent from the registered name’s record. A needle is never invented by diffing two responses: the wrong needle turns a registered domain into an AVAILABLE, which is the one answer ds must never give. A TLD whose server cannot be shown to tell the two names apart is left out of the file rather than guessed at, and ds falls back to asking IANA at runtime.

scripts/whois-report.tsv records the verdict and the reason for every TLD considered, so a rejection can be looked up rather than wondered about.

./scripts/refresh-whois.py all        # harvest, verify, rewrite whois.json
cargo build --release
./scripts/refresh-whois.py verify     # re-check the table through ds itself

The run is paced to be a good guest: queries to one registry are serialised with a gap between them, keyed on the address the server resolves to rather than its name, because hundreds of whois.nic.<tld> aliases sit on a handful of shared back ends. Expect it to take hours.

These pages are generated from the docs in the repo, so they never drift from what ships with the binary.