Usenet, Bulletin Boards, and the End of the Editorial Gatekeeper

A network that had no office to raid

In 1979, two graduate students, Tom Truscott at Duke and Jim Ellis at the University of North Carolina, sketched a way to link Unix computers over ordinary telephone lines using a file-transfer tool already installed on most of the machines around them, the Unix-to-Unix Copy protocol, known as UUCP. The system they built on top of it, Usenet, was running by 1980. What they had assembled, without quite setting out to, was a publishing network with no publisher.

Every mass medium before it had a chokepoint a government or a company could find and control. A newspaper had a press, a warehouse, and a masthead someone owned. A radio station had a transmitter and a license issued by a national regulator, the kind of license the United States began requiring of every broadcaster once the Federal Communications Commission was created in 1934. Usenet had none of that. It was distributed across a large, constantly changing set of independently operated news servers, each one owned and run by whoever chose to run it, none of them requiring a dedicated administrator or a paid hosting arrangement, exchanging messages with their neighbors through peer to peer transfers called feeds until a post had travelled from one machine to thousands.

The first newsgroup, net.general, set the pattern the rest of the network followed. Individual servers were under no obligation to carry any specific newsgroup, so the operator of a single machine could refuse a category of discussion on that machine, and nowhere else. That is a strange kind of editorial power: total inside one node, worthless everywhere else in the graph. No server had to say yes on behalf of the network as a whole, because there was no whole to ask permission of. There was only the next machine down the phone line, and its own operator's own decision.

That single design choice, thousands of independent local editors and no global one, is the thread running under everything that follows. It is what let Usenet destroy the old economics of gatekept publishing, and it is also what let the network's first advertiser exploit exactly the gap that made it possible.

What a press used to cost

Reaching a mass audience had always required capital or a license, usually both. Movable-type printing spread across Europe from the middle of the 1400s and still needed a press, plates, paper, and a way to move finished copies to readers. A broadcaster needed a transmitter and a slot on the spectrum a government agreed to grant it. Usenet needed a Unix account and a modem. There was no press run to fund, no truck to load, no tower to build, and no application to file with a regulator before a message could reach a reader on the other side of the world.

The effect was a collapse in the price of reach, not a small discount on it. Once a machine had a working feed, publishing to that feed cost close to nothing beyond the connection itself, so the set of people who could address a worldwide audience widened overnight from a small list of institutions with capital or state approval to anyone with a terminal and a phone line. Because no publisher pre-screened what went out, the norms an editor would ordinarily have enforced had to be invented by the people using the network, in public, as they went. Usenet is credited with giving the language much of the vocabulary it still uses for that kind of self-policing: the FAQ, the flame, the sockpuppet.

None of that vocabulary described a business model. It described a commons figuring out, with no owner and no editor, how to behave. For the better part of fifteen years that worked as a functioning, if noisy, form of self-governance. It also left the network with no mechanism, and no one whose job it was, to say no to anyone who wanted to use its free reach for something other than conversation.

The first spam, and the money in it

On April 12, 1994, Laurence Canter and Martha Siegel, a pair of immigration lawyers, posted an advertisement titled "Green Card Lottery: Final One?" to at least 6,000 Usenet newsgroups simultaneously, most of which had nothing to do with immigration law. It is widely regarded as the first mass commercial spam campaign run on the network, and it worked precisely because Usenet had no publisher positioned to refuse it. There was no editor at the network level who could have said no. There was only the eventual reaction of thousands of individual readers, each on their own machine.

That reaction came fast, and in a form the network had not needed before at that scale: automated "cancelbot" software, run at real scale for the first time to mass-delete the pair's posts across every server willing to honor the deletion. Canter and Siegel's own internet service provider fared worse. Its mail servers crashed repeatedly under the volume of complaints, and the provider terminated the couple's account within days. Judged as a test of whether an open network with no owner could defend itself, the answer was clear: it could, though not gently, and through no authority higher than its own users and their tools.

Judged as a business decision, the campaign looked different. It reportedly generated between $100,000 and $200,000 in business, a figure drawn from period accounts rather than an audited record and best read as a contested estimate. Contested or not, the direction of the finding held: bulk unsolicited posting to a network with no gatekeeper and no per-message cost could be immediately and directly profitable. That is the template the commercial email spam industry would go on to scale for the following three decades. The same absence of a publisher that let two graduate students give a global audience to anyone with a modem also gave that audience, at no charge, to anyone willing to spend an evening addressing an advertisement to six thousand rooms at once. Spam was not a flaw discovered later. It was latent in the architecture from the day the architecture removed the one person whose job it would have been to say no.

No server to seize

The same feature that broke the economics of gatekept publishing, no single chokepoint, also broke a government's usual lever over a medium: control of its infrastructure. A state could revoke a broadcast license, seize a press, or lean on a phone company that owned the wires a signal travelled over. Usenet was harder to reach that way because it did not run through one company or one country's infrastructure. It ran through a very large number of independently owned machines, each connected only to whichever neighboring machine happened to answer its call.

FidoNet made the pattern visible at the edges of the network before most people had internet access at all. In the early 1990s, before internet access was broadly affordable, a large share of Usenet's content reached ordinary users indirectly, through FidoNet, a separate network of dial-up bulletin board systems that relayed long-distance and worldwide discussion on its own, outside the control of any single national telecom monopoly. A hobbyist BBS operator in one city could exchange a night's worth of messages with a counterpart in another country over a local phone call and a modem, then pass those messages on to the next board down the chain, and the discussion crossed a border without a treaty, an export license, or a request to either government involved.

That was not incidental to Usenet's design, it was built into the core mechanism: any content posted to a newsgroup automatically proliferated to any peer server that carried that newsgroup, wherever that server happened to sit. Content posted in one country did not need a cross-border commercial agreement or a government sign-off to appear in another; it simply moved to the next machine down the line the way it moved to every other machine down the line. Information moving across borders on a topology no single government could unilaterally shut down is a direct ancestor of the jurisdictional headaches that would come to define control of the open internet: whose law applies to a server room in a country neither party to a dispute lives in, and what a government can actually do about content when there is no single machine left to seize.

The gatekeeper Usenet removed, and the one that replaced it

Usenet answered one question completely: who may publish to a global audience. The answer, after 1980, was anyone with a working feed. It never had to fully answer a second, harder question, because for most of its life nobody was reading at global scale on a reader's behalf: who gets read. That second question is the one running through the current chapter of the same argument this series is built on, that control of the dominant medium of an age shapes its economy and its politics. The system that now decides who gets read at scale is not a newsgroup server. It is an AI engine assembling an answer to a question a buyer asked, and the businesses it names are the ones whose information it could find, parse, and trust enough to cite.

Usenet's history is a reminder that removing the gatekeeper from the right to speak does not remove the gatekeeper from being heard. It only moves that second job somewhere else, first to search engines ranking pages, and now to systems deciding, one answer at a time, which of the millions of things that could be said actually gets repeated back to the person who asked.

Usenet's fifteen years show both halves of that pattern at once. It liberated the right to publish from capital and from license, which is why a discussion culture with no publisher could still produce the working vocabulary the rest of the internet later adopted for its own self-governance. It also demonstrated, inside its first decade and a half, that removing a gatekeeper does not remove the need for governance, it only relocates the cost of that governance onto whoever ends up running the cancelbot, crashing under complaint volume, or, in the current chapter, deciding which sources an answer engine is willing to trust.

What the record does and does not show

The claims in this account rest on two kinds of source. The structural facts, no central server, independently run peer to peer feeds, per-node discretion over which newsgroups to carry, are well established and corroborated by the technical record of how the network was built. The Canter and Siegel campaign and its cancelbot backlash are also well corroborated, reported at the time and revisited since. The dollar figure attached to that campaign is not: the range of $100,000 to $200,000 comes from period business reporting rather than an audited account, and should be treated as a contested estimate of what a gatekeeper-free network's first advertiser made, not a settled number.

The geopolitical reading carries a caveat of its own. A network with no central server was hard for any government to shut down at its source, but that did not make it impossible for a government to restrict what its own citizens could reach; a state could still act on the internet providers and BBS operators inside its own borders, even if it could not reach into a server sitting on someone else's soil. What Usenet demonstrated was narrower and still significant: no single government could unilaterally close the network for everyone else on it, a different claim from saying no government could touch it at all.

The parallel to today's answer engines is offered as a reading, not a proof. Usenet's per-node carry-or-refuse decision and a modern AI system's judgment about which sources to trust and cite are not the same mechanism. One is a binary local choice made by a server operator; the other is a ranked, probabilistic judgment made across a whole corpus at once. What they share is a position in the argument this series is built on: whoever controls the layer that decides what gets seen, in any age, captures a share of the economics and the politics that layer touches. That is the pattern under study across this series, not a claim that Usenet caused, or even resembles in mechanism, the systems running today.