The Hook: A Hacker News story slipped through the feed digest — ".name Termination" — two thousand upvotes, hundreds of comments, everyone in shock. It smelled like the typical internet drama of "corporation screws everyone over," but when checked for repeatability, the hook turned out to be virgin territory: not a single previous investigation had touched this topic. And it's worth it, because this isn't just "someone's website getting taken away." This is the closure of an entire architectural DNS layer that existed for a quarter century, formalized with a couple of PDF forms that engineers filled out with three-word answers.
Neil Fraser is no small-time complaining blogger. He's an engineer who worked at Google for almost twenty years (creator of Blockly — that visual programming library half the planet's children first encountered code with), and now at New Voice in Switzerland. In January 2002, literally days after the new .name zone launched, he registered neil.fraser.name — the domain became his website, email address, and API server for decades to come. When his daughter was born, he registered beverly.fraser.name — by his own account, within minutes of delivery. Beverly's domain is alive until 2030, Neil's own domain is paid through 2036. Both exist in Verisign's RDAP registry right now, status "active". What will kill them isn't expiration — the zone operator itself is going to kill them.
On April 15, 2026, Verisign filed two requests with ICANN under the RSEP (Registry Services Evaluation Policy) procedure: the first — "termination of third-level domain registration in .name" (RSEP 2026013), the second — "discontinuation of email forwarding service" (RSEP 2026014). The wording in the application is a model of minimalism, for which registrars justifiably caught contemptuous comments on HN:
Yet the request itself describes that "existing third-level registrations will be deleted." The questionnaire was filled out as if they already knew no one would seriously check the answers — which proved true.
On May 7, 2026, ICANN approved the request. On July 28 — issued a confirmation letter (instead of a contract amendment, because the agreement with Verisign happens to be on the verge of renewal). ICANN laws don't provide for a public comment period for RSEP approval at all — the BAMC speaker would later politely record this in an official document. Users received notification about the fate of their paid domains from... their own registrars, when the procedure had already been launched.
To appreciate the scale, we need to go back to 2000. ICANN conducts the first round of new gTLDs: .aero, .biz, .coop, .info, .museum, .pro, and .name. London-based Global Name Registry (GNR), owned by Nameplanet, goes for .name — a company that by then had already collected 700,000 users of a service giving people email like name@surname.com. GNR's idea was fresh for an era when .com was already gasping: internet for people, not companies.
But GNR's engineering solution was truly non-standard. The problem: few surnames, many people. If you let people register smith.name directly, one lucky person buys the Smith surname — and what do the other millions do? GNR's solution: the zone is rigidly structured at the third level. The surname (smith.name) always belongs to the registry, and people register john.smith.name, mary.smith.name — as many Johns and Marys per surname as needed. Second-level registration is directly prohibited in the original Registry Agreement. Plus a parallel email service: firstname@lastname.name — an address doesn't require owning a domain at all, the registry holds the DNS records. And this mechanism went into production on January 15, 2002 — with a playlist of "donald.duck.name" and "mickey.mouse.name" in the first WHOIS batches.
The launch story is striking: 60,000 pre-registrations by day one, the British government with the e-commerce minister personally supporting the launch, the first "owner" in the zone's history — newborn Katie O'Rourke, whose parents registered katie.orourke.name at her birth. GNR predicted "tens and hundreds of millions" of names. One of the first registrations — Fraser — is recorded in RDAP data: January 23, 2002, 14:41 UTC. Today this domain no longer resolves for anyone among the "classic" names of that era: katie.orourke.name, john.smith.name — "No match". The first launch characters already died before the official execution date.
Peak of the .name zone — 232,408 domains in 2012 (Verisign data in ICANN public comment). By 2026 — around 96,000. Actually resolving are about 63,000 A records. And of those, actual third level — 22,288 domains: Verisign told ICANN rounded "about 22,000," and the exact figure was calculated by Doychin Spiridonov, employee of Bulgarian registrar Dom.bg, independently from the zone file, because no one published official breakdown of "how many of the 96,000 are alive and who exactly will be affected." The level of actual death: among the 22,288 living third levels — for example, 37 Kevins (domains like kevin.*.name), which Kevin Murphy himself pointed out in the Domain Incite headline — a publication whose author shares that name. This isn't exaggeration, it's a cited figure from an industry publication.
Three causes of death, all banal to tears:
firstname.lastname.name format broke familiar EPP flows; a significant portion of registrars only sold yourname.name. Unofficial lists compiled by enthusiasts of registrars capable of handling third level were described for years with the formula "very hard to find.".me (over a million domains, slogan "pronoun marketing"), link-in-bio platforms, and plain old .com. The firstname.me format is shorter, sweeter, and doesn't require explaining to users that the dot between first and last name isn't a typo..com/.net and ~600 billion DNS transactions per day. For an entire category of domains and people — "deal amount not disclosed."In 2009, Verisign, renewing the agreement, allowed second-level registration directly — effectively admitting the original third-level philosophy hadn't taken off. At the same time, in 2012, the zone got DNSSEC. The zone didn't die from technology. It died because the idea of a "personal domain" was ahead of its time: before the social media era, people really needed a "page of their own name," but when social networks took that niche for free, .name was left without a reason to exist.
The "shutdown of an unpopular service" itself is normal engineering hygiene. The plot is more dramatic, and here's why.
After third level dies, freed surnames (smith.name, fraser.name, and 22 thousand other second levels that Verisign hadn't let anyone register for a quarter century) become available for registration anew. In the RSEP about how exactly — drop, auction, priority for old owners — nothing is said. The Chinese section of the giant domain speculator market is already writing instructions "sell your third levels, prepare to catch surnames a year after mail service closure" (portal vname.com estimates a one-year window). Essentially the largest surname auction in history is launching — with half of them already having known new "owners" by the first person whose name matches.
And here begins what Fraser writes about calmly, but which looks like an announcement of catastrophe. Third-level .name domains aren't just websites. They're twenty-five years of an unknown number of address-linked logical entities:
No one knows how to enumerate all the places this was used. And when fraser.name is freed and goes to someone else, the new owner will be able to recreate neil.fraser.name and gain: DNS control, email interception (anyone who wrote to that address), administrative access to IoT devices, and — most piquant — the ability to commit code in his name. This is a ready recipe not just for "stole a website," but "stole digital identity with a transitive dependency list." One engineer. But there are twenty-two thousand like him, and most haven't even read the RSEP.
Not everyone accepted this silently. Doychin Spiridonov (better known in the domain industry under the handle bgc, Bulgarian registrar Dom.bg) filed with ICANN a Request for Reconsideration 26-2 — a reconsideration procedure, formally the only channel to challenge headquarters decisions. His claims are precise and surgical: approval was made in opaque consultations between Verisign and ICANN, conducted in closed mode in October–December 2025; approved forms contained "materially inaccurate, incomplete, and misleading information"; and — central thesis — Verisign's statement "no effect on the life cycle of domain names" is simply false, because a domain's life in ICANN standard begins with registration and ends with expiration, not premature killing from the center.
On August 24, 2026, BAMC (Board Accountability Mechanisms Committee) recommended denial. Not because the facts weren't proven — but because formally ICANN is right: by RSEP definition, a domain's "lifecycle" begins with registration and ends with deletion, and deletion, including premature deletion, is a standard lifecycle stage, so deleting a domain "doesn't change the lifecycle." Ombuds reports with wording that makes you want to lay down the monitor: "the answers [were] resoundingly clear." Public comment for RSEP approval isn't required. And grandfathering through paid dates or refunds — "beyond the scope of the reconsideration process and relates to registrar-registrant contractual relationships, to which ICANN is not a party."
This is a rare case where legal casuistry and engineering meaning diverge, and the lawyer wins formally: neither ICANN nor Verisign is lying — they're just moving terms. And 22,288 people, many of whom received their paid future (Fraser's — until 2036), end up between the chair of "registrar is obligated, by contract, to fulfill your wish at their discretion" and the chair of "ICANN is not a party to your contract."
Verification of Verisign's history at the time of writing showed that internet community trust in it isn't unfounded bias from an engineer whose domain is being taken. In 2003, Verisign deployed Site Finder: a wildcard record for all unregistered .com/.net, redirecting every typo address to the company's advertising portal. This captured traffic from millions of daily typos and broke spam filter behavior, SMTP clients, network printers. ICANN had to send an official letter demanding removal of the wildcard — the service was shut down on October 4, 2003. In February 2004, Verisign sued ICANN, claiming it exceeded its authority; the lawsuit was dismissed in August 2004, the case ended with a settlement in 2006 with a new Registry Agreement.
Actually, against this résumé it's unsurprising that Fraser in 2002 chose .name precisely because "GNR — definitely not Verisign." Getting into the era of short combs, Verisign eventually became the operator of the very zone Fraser tried to escape from. The computer writing history is sometimes more ironic than any screenwriter.
Conclusions
Three layers that stick in your head here.
First — about "standards vs reality." Formally ICANN is impeccable here: procedure followed, RSEP is consensus policy, reconsideration mechanism — proven, BAMC response — exemplary legal text. Engineering-wise everything looks like the cheapest API deprecation in the world: not "migration plan for 22,288 users," not grandfathering, not buy-back, but one line in a PDF ("existing third level domain names will be deleted") and a letter to registrars 90 days out. In a world where one of the first global identity templates on the internet is a domain in a zone created "for people" — ultimately, no one has a migration plan in case "the industry stands idle." For an industry where phrases "architectural stability" are written in capital letters, this looks indecent.
Second — about how "personal space" on the internet isn't property space at all. A domain paid through 2040 is a 14-year-forward lease, with the landlord's right to demolish the building. Fraser honestly writes the address was registered "as stable presence" — and didn't attach significance that beneath it stands not a contract with ICANN, but a contract with a zone operator that itself will one day become an unprofitable segment of someone else's portfolio. Any "eternal" digital thing — domain, account, certificate — in reality has a single point of failure with the right to prematurely change conditions. If Verisign had obligated owners to relocate and bought back surnames for nominal — the story would look different. Instead, a hunt for surnames opens, where the owner of a 25-year digital asset has no priority, no buyback right, not even the right to know the exact auction date.
Third — historical rhyme. In 2002, .name launched as "internet for people": with a minister, with newborn Katie in the maternity ward, with "a clever idea to protect millions of Smiths from greedy surname buyers." Twenty-four years later, the first floor of that very idea will be demolished for administration simplification. GNR's 2000 forecast — "tens and hundreds of millions of domains" — came true at 0.01%. Katie O'Rourke, by the time you read this, is already an adult, and her domain, judging by the registry, has been dead for several years. This isn't a tragedy on the scale of "NASA couldn't find $2M for a unique mission" — there at least there's a conflict of budget and science. Here the conflict lies deeper: 22,288 forward-paid leases interfere with simplifying one legacy service in someone else's portfolio, and so their existence is simply canceled retroactively. One decision writes history, the second — just a line in the books.
A quiet morning in February 2027: Fraser's server returns its last 404, IoT sensors in his home stop knocking on DNS. Outside the window — an ordinary Tuesday, the internet hasn't changed a bit. Just 22,288 fewer names in the .name zone, and no one except their owners and a couple of domain publications noticed how an entire architectural solution simply ended.