Summary

  • The .kids registry agreement makes child-appropriate content a condition of registration, covers the domain name itself, and permits suspension or cancellation. Its stated enforcement route begins at an online complaint portal while expressly avoiding prescreening.
  • A University of Twente study presented at RIPE 93 tracked 1,427 daily snapshots through March 2026. Its slides report a one-day median to the first external abuse flag, 34 days of median flag persistence, and 66% of flagged domains still active. Those clocks leave one operational question unanswered: when did a usable report reach the registry, and what happened next?

The sign is lovely. Who reads the inbox?

A children's garden needs a gate. A domain registry can put a bright sign on one, sell the addresses inside, and call the result a protected neighbourhood. The hard work begins after someone reports the broken lock. In the .kids namespace, the sign is unusually specific. The ICANN registry agreement says content and services under .kids must be suitable for people under 18. It names gambling, illegal drugs, pornography, violence, alcohol, tobacco and criminal activity. Even the domain string counts as content. A violation can ground suspension or cancellation.

That is a much stronger promise than a cheerful suffix. Specification 12 also tells the operator how the promise is to be policed: an online complaint portal starts a takedown decision process under a “Protection Scheme”. The same text rejects prescreening and unnecessary censorship, citing children's freedom of expression and access to information. Those two choices create a real operating design. The registry will let names in, hear complaints, judge them, and act. A gatekeeper who does not inspect every visitor needs an excellent complaints desk.

On 27 October, University of Twente researchers brought an original .kids study to RIPE 93. Their nine-slide presentation describes 1,427 daily snapshots from May 2022 through March 2026, combining zone measurements with ICANN transaction reports and outside classification and abuse signals. More than 30% of names were inactive, parked or placeholders at any given point in their account. More than 85% of registrations used the minimum one-year term. A name sold is a registration; it is not a child-friendly service delivered. The nursery looks busier on the sales ledger than it does when the lights come on.

The most useful slide is about time. The researchers report a median of one day to the first external flag, a median of 34 days during which flagged domains persisted, and 66% of flagged domains remaining active in their measured frame. The figures describe the study's detection and zone observations. The agreement describes a complaint-triggered decision process. Between those two records sits the missing receipt. The public originals here show no per-domain timestamp for an accepted complaint, review, decision, suspension, appeal or restoration. A flag may never have entered the operator's portal.

A complaint may have been found inapplicable. The way to find out is to publish the handoff, not to call the 34 days a registry response time.

That distinction matters more than a dramatic count of bad names. An external scanner can tell an operator where to look; it cannot adjudicate a contract. A registrar can sell a domain; it cannot make the registry's community promise disappear. ICANN can hold the registry to its agreement; it needs an observable trail when the dispute is about execution rather than the wording of Specification 12. The original presentation and the event abstract even use different counts of apparent agreement violations. The safety question survives that small numerical quarrel: what did the registry learn, when, and by which route did it decide?

An operator could answer with a compact public ledger: complaint receipt date, category, triage result, decision date and action status, aggregated enough to protect complainants and children. It could separate a named-content allegation from a phishing flag and a parked page. It could show how often a case is cleared because a site changed, how often a domain is suspended, and how long a review waits. None of that demands automatic takedown from a machine-generated alert. It asks the institution selling a safety promise to account for its own desk.

The .kids contract gives the desk legal furniture: rules for the label, grounds for removal, and a process that takes speech rights seriously. The RIPE measurements look through the front window and find many empty rooms and some dangerous ones. The next governance test is prosaic enough to be credible: open the complaint register, show the route from signal to decision, and let the dates speak before the brochure does.