On June 21, the Human-Centered Computing Foundation published its vision for .self — a proposed top-level domain designed to serve as a permanent address for self-hosted infrastructure. The pitch, posted to HCCF’s site and discussed widely on Hacker News, is earnest and detailed: a domain that maps to a person, not a platform; a namespace that persists even as hosting providers come and go; a small piece of internet real estate you actually own.
The timing is no coincidence. ICANN’s 2026 round of new generic top-level domain applications opened on April 30 and runs through August 13. An application costs $227,000 before legal and technical fees. HCCF is asking ICANN to classify .self as a community-based TLD — a designation that, if granted, would let it bypass the auction process that turns contested strings into a bidding war between deep-pocketed portfolio registries.
It is a thoughtful proposal from people who clearly mean well. It will almost certainly fail to deliver what it promises, and the reason is not ICANN bureaucracy. It is physics — or at least economics.
The Last Mile Has a Landlord
The vision of .self is seductive because it solves for identity. Your domain is yours forever; your services live behind it; the address never changes. What it does not solve is connectivity.
Self-hosting requires a connection to a last-mile network — and in the United States, 83 million households have exactly one or zero options for broadband above 25 Mbps, according to the FCC’s most recent deployment report. If you have a choice, it is likely between one cable operator and one legacy telephone company that stopped investing in copper maintenance a decade ago. Neither wants you running services from your home. Their terms of service generally prohibit it on residential plans, and their upload speeds — still asymmetrical across most of the country — make it impractical even where it is technically permitted.
A domain name does not peer your traffic. A domain name does not negotiate with Comcast. alice.self resolves to an IP address that sits behind a carrier-grade NAT or a modem that reboots when the firmware update arrives at 2 a.m. The address is permanent; the path to it is not.
“The domain is the easy part,” as one network engineer at a mid-tier ISP put it when I described the proposal. “I can give you a /29 in my sleep. What I can’t give you is a business-class SLA on a residential price point, and nobody’s re-architecting the DOCSIS plant so you can run a Matrix server.”
A Brief History of Domains That Were Going to Change Everything
New TLDs have a near-perfect track record of disappointing their idealists. .org was supposed to be for nonprofits. .info was supposed to be for informational resources. .name was supposed to be for individuals. .eco was supposed to signal environmental commitment. Each of them either became generic inventory for domain investors or drifted into irrelevance.
The mechanism is consistent: a TLD launches with a compelling narrative, early adopters rush in, and then the economics of renewal take over. A registry needs revenue. Premium names go to auction. Domain parking firms snap up anything with search volume. Within three years, the namespace looks indistinguishable from .com — just smaller, with less trust from browsers and email filters, and a renewal fee that crept up 40% because the registry’s venture backers need a return.
.self would enter this ecosystem with a built-in tension. Its value proposition depends on long-term stability — the domain as permanent anchor. Its funding model, unless HCCF has an endowment it has not disclosed, depends on renewals. Those two things pull in opposite directions.
Who Actually Gets Liberated
ICANN’s community priority evaluation process is designed to favor applicants who genuinely represent a defined community and will operate the TLD in its interest. It is also a process that requires lawyers, consultants, economists (for the community scoring analysis), and a non-refundable application fee. The application window is 105 days. The evaluation takes months. Objections and disputes add more months. A contested string can spend two to three years in procedural limbo before a single domain is delegated.
This is not a critique of ICANN’s process specifically; it is a description of any regulatory gate with a $227,000 cover charge. The people best positioned to navigate it are not the people running a home server off a Raspberry Pi. They are the professionals who bill by the hour to help other people navigate it.
HCCF may well clear the community bar. It is a real organization with relevant expertise. But the path from approval to delegation to actual adoption runs through a series of decisions — about pricing, about abuse mitigation, about which registrars get access — that will be made in conference rooms and Zoom calls, not in the spirit of the original Hacker News thread. The consultants will do fine. The question is whether anyone else notices a difference.
The Thing That Would Actually Help
If the goal is to make self-hosting viable for normal people, the bottleneck is not the DNS. It is the fact that running a service from your house violates the acceptable use policy of the one ISP you can buy from, and even if it didn’t, your upload pipe is 20 Mbps on a good day.
Fixing that requires a regulatory intervention — something closer to common-carrier obligations for last-mile networks, or municipal fiber builds that treat symmetric connectivity as infrastructure rather than a premium upsell. It is a heavier lift than a TLD application. It is also the thing that would actually make the .self vision work.
The .self proposal is elegant, carefully argued, and aimed at a real problem. It is also a software solution to a hardware problem — a familiar instinct in the engineering communities where it originated. Domains are nameplates. The question is whether there is a door to put them on.