The Domain Name System and the Contest for the Internet's Address Book

A phone book that only one office could print

The Domain Name System exists because numbers are hard to remember and names are not. Every device on the internet is reached by a numerical address, and the System's job, running as an essential and continuously operating part of the network since 1985, is to translate a human-readable name into that address on request (Wikipedia, "Domain Name System", 2026). Before it existed, the whole network shared one flat text file listing every host, a scheme that worked while the internet was small and stopped working once it was not. The System replaced that file with a hierarchy: a root, then top-level domains such as .com and .edu, then the names underneath them, each layer delegated downward from the one above.

A hierarchy needs a top. Someone has to hold the record that says which top-level domains exist and which servers are authoritative for each one, and that record is the root zone. For the system's first thirteen years, the person who held it was Jon Postel, a computer scientist who had been present at the internet's earliest technical decisions and who ran the Internet Assigned Numbers Authority, the body responsible for the DNS root and for allocating the numerical addresses underneath it, from its inception until his death in October 1998 (Wikipedia, "Jon Postel", 2026). He was informally known during his life as the "god of the Internet," a title that reads as hyperbole until it is measured against what the role actually controlled: the master list of every name the network would recognize as valid.

That is the plain mechanism at the center of this article. A name means nothing on a network until something authoritative agrees to route to it. For the internet's formative decade, one man's server held that agreement, and the government that funded his work held the only formal claim on him.

It is worth being precise about what "the root" actually is, because the phrase can sound more abstract than the object it describes. The root zone is a file, kept on a small set of servers, listing the top-level domains the System will recognize and the addresses of the servers authoritative for each one. It is not large, and it is not, on its own, technically complicated. Its importance is entirely a function of position: every other lookup in the System eventually traces back to it, so a change to that file, adding a top-level domain, removing one, or redirecting where one points, changes what the entire network can reach. Holding the pen on that file was, functionally, holding a veto over what could exist as a recognized address on the internet.

The economic thread: a name becomes a precondition

The System was built as a convenience, a lookup service to spare users from memorizing strings of digits. It did not stay a convenience for long. As the internet opened to commercial use in the early 1990s, having a name under a recognized top-level domain, above all .com, stopped being optional. A business without a domain name could not be reached by the addressing scheme the rest of the network used to find anything, which meant it effectively did not exist to a customer arriving by browser rather than by word of mouth. The precondition was not a webpage or a product; it was a line in the root's delegated records.

Scarcity does what scarcity always does to a namespace with real estate value: it creates a market and then a fight over the market. Because a .com name was unique, first-registered, and increasingly necessary, names themselves acquired resale value independent of anything built on them, and the same period produced the internet's first monopolistic naming disputes, arguments over who had the right to hold a name that a trademark holder, a later entrant, or a speculator all wanted at once. None of that dispute machinery existed when the System was designed in 1985. It was a direct, unplanned consequence of turning an addressing convenience into the entry ticket for a commercial network.

The commercial opening itself moved in stages, and the dates matter because they show how quickly the precondition hardened. The National Science Foundation Network, a U.S. federally sponsored program that formed a major part of the internet's backbone from 1985 to 1995, initially restricted use to government agencies and universities, and the first commercial internet service provider did not appear until 1989 (Wikipedia, "NSFNET", 2026). The network's own Acceptable Use Policy, drafted between 1988 and 1990, formally barred commercial and for-profit traffic from the taxpayer-funded backbone, though in practice the ban proved largely unenforceable as companies routinely used the network regardless (Grokipedia, "Acceptable use policy"; Technics History, "Internet Ascendant, Part 2," 2020). By 1991 the NSF had dropped the access restrictions outright, and in 1992 it announced its intention to phase out federal support for the backbone entirely, actively encouraging private companies to build competing commercial backbones of their own (Wikipedia, "NSFNET", 2026; NSF.gov, "Birth of the Commercial Internet"). Federal funding for the NSFNET backbone formally ended on April 30, 1995, completing the handover from a government research network to a privately operated commercial one (Technics History, "Internet Ascendant, Part 2," 2020).

So the sequence runs: a network built for researchers, opened to commerce faster than its own rules could keep pace with, handed off its physical backbone to private operators inside a decade, while the naming layer that made any of it findable stayed under a single individual's informal stewardship the whole time. The economy caught up to the network before the governance did.

Registration itself, in this period, ran through a small chain of delegated authority rather than a free-for-all. Names were assigned first-come, first-served under the emerging registration practice that grew out of the root's delegated structure, with essentially no verification that the registrant held any trademark or business right to the string being claimed. That single design choice, register first and sort out rights later, is what made the domain a speculative asset in the first place. A word or a company name had no gatekeeper checking prior claims at the moment of registration, so the fastest typist, not necessarily the rightful trademark holder, got the name, and the argument over who deserved it happened only after the fact, in courtrooms and in the earliest domain-dispute policies that ICANN would later formalize.

The policy scaffolding behind the handoff

The privatization of the backbone was not spontaneous. It followed a policy track that had been laid years earlier. The High Performance Computing Act of 1991, sponsored by then-Senator Al Gore and signed into law on December 9, 1991, allocated roughly 600 million dollars to fund the National Research and Education Network and helped set the policy groundwork for the network's eventual public and commercial expansion (Wikipedia, "High Performance Computing Act of 1991", 2026). The legislation is often remembered, loosely, as an early government bet on what the internet could become; its more precise legacy is that it funded expansion while the addressing authority underneath that expansion remained outside any statute at all.

That asymmetry is the point worth holding onto. Congress could authorize hundreds of millions of dollars for network research and infrastructure, and an agency could fund and then divest a backbone on a fixed schedule, all through visible, dated, legislated steps. The root of the naming system that made the whole thing usable had no equivalent instrument. It was a research contract and a trusted individual, not a law.

The geopolitical thread: infrastructure with one government's name on it

Here is the fact that gives this era its geopolitical weight. The United States government funded the research programs that built the internet's early infrastructure, and a single U.S.-based individual administered the authoritative root of the entire network's naming system, for over a decade, with no binding international oversight. Other governments, including close allies of the United States, had no formal seat at the table that decided which top-level domains would exist, which country codes would be recognized, or who would administer them. The System was, in that period, a piece of global public infrastructure operating on the informal trust extended to one man and the funding relationship he held with one country's government.

This was not a secret or a scandal at the time. It reflected where the technology had actually been built, and Postel's stewardship was, by the accounts of people who worked with him, careful, and unusually resistant to self-interest for someone holding that much informal power. But informal trust is not the same as formal accountability, and a concentration of infrastructural authority does not need to be abused to be contested once the infrastructure becomes indispensable to nations that had no hand in designing it. As the internet moved from an American research project to a genuinely global commercial and social network through the 1990s, the gap between who used the System and who governed it became harder to defend on its own terms.

The record of that contest played out slowly and mostly through institution building rather than confrontation. Its first formal resolution arrived just past the close of this era: the Internet Corporation for Assigned Names and Numbers was officially incorporated in California on September 30, 1998, explicitly to institutionalize DNS root and naming governance that had previously rested on Postel's individual stewardship, with the technology entrepreneur Esther Dyson serving as its founding chairwoman (Wikipedia, "ICANN", 2026). Incorporation, though, did not mean independence from the United States. ICANN's contract to perform the IANA stewardship function was held with the U.S. Department of Commerce's National Telecommunications and Information Administration, and that contract did not formally transition to the global multistakeholder community until October 1, 2016, meaning the U.S. government kept a direct contractual link to the internet's naming authority for roughly two more decades after the window this article covers (Wikipedia, "ICANN", 2026).

Read plainly, the geopolitical arc runs longer than the technical one. The System went from one American's server to an American nonprofit under an American federal contract to, only in 2016, a genuinely international arrangement. Every step in between was a government or a set of governments negotiating, in slow motion, over a piece of infrastructure that had been built and settled before most of the negotiating parties had any power to shape it.

Every medium liberated and concentrated at once

The System's double effect cuts both ways: it liberated as plainly as it concentrated. Before it existed, using the network at scale required memorizing or looking up numerical addresses, a barrier that kept the network's practical reach narrow even among people with formal access to it. Naming took that barrier away. Once a name could stand for an address, a much larger population of people and, soon, businesses could use the network without technical training, and a name became something a person could build a memory, a brand, and eventually a livelihood around. That is a real liberalization, and it is why the namespace became valuable enough to fight over in the first place.

The same mechanism that liberalized access concentrated authority over the thing being accessed. A namespace that anyone can use but only one party can authoritatively define is not neutral infrastructure; it is a chokepoint with a friendly face. The individual who ran the root in this era used that position with restraint, by most contemporaneous accounts, and restraint is precisely the quality that a durable governance system cannot depend on indefinitely, because it is a property of the person holding the office rather than of the office itself. The move toward ICANN, however incomplete and however long delayed, was the recognition that a resource this consequential needed rules that outlasted any one administrator's good judgment.

The two effects are not a contradiction to be resolved. They are the same fact, seen from two audiences: a business owner registering a name and a foreign ministry watching who holds the root are looking at the identical piece of infrastructure and drawing opposite conclusions about what it means for their position, and both conclusions are correct.

The line to the present, where it earns it

The chokepoint this era built was over naming: who could be found under a name at all. The chokepoint of the present era, the one this publication studies, is over answering: whether an AI system, deciding who to name in a response, can find, read, and trust a business enough to cite it. The two are not the same mechanism, and it would overstate the case to claim a straight line of cause between a 1985 addressing protocol and a 2026 answer engine. But the underlying pattern repeats, and it is fair to name it.

In both eras, the layer that decides who is visible to everyone else was built and settled by a small number of actors before the businesses depending on it had any say in its rules, and the businesses that adapted earliest to the new layer's actual mechanics, registering a defensible name in the 1990s, or being legible to an answer engine now, captured an advantage that later entrants had to buy back at a markup, in cash for a resold domain then, in visibility work now. The System's history is a reminder that a naming layer is never just plumbing. It is the address book the rest of the economy has to consult, and whoever writes the address book sets terms nobody else voted on.