An Ethnographic Study of the First Digital Commons

Between 1979 and roughly 2010, the largest sustained public conversation in human history took place on a network that most people today have never heard of. Usenet — a decentralized system of newsgroups running over ordinary phone lines — hosted millions of people arguing, teaching, grieving, joking, praying, and organizing across every topic human beings have ever cared about. It was not owned by anyone. It had no central server, no algorithm, no advertising. It simply existed, and people filled it with what mattered to them.

At its peak, Usenet carried more than a billion messages in fifteen thousand newsgroups. It hosted the first online communities for pagans, Gnostics, Buddhists, chaos magicians, Sufis, and Hindus in diaspora. It is where internet culture — the meme, the FAQ, the flamewar, the troll, the regular — was invented. It is where the architects of 4chan and Reddit and Wikipedia learned to think about online community. It is where a generation of practitioners found their traditions, argued their heresies, and became who they were.

Most of it is gone. What remains is held in incomplete archives, maintained by volunteers, gradually degrading. This history was written to accompany the New Tianmu Anglican Church's ongoing project to preserve the religious and spiritual content of the Usenet newsgroups before that content is lost entirely — to give the archive its proper context, and to explain why the conversation it preserves was worth having.

What follows is an ethnographic study of a lost civilization. It treats Usenet with the seriousness that historians bring to any culture that invented something important and then disappeared — which is what Usenet did.

Before the Net — The Prehistory of Online Community

When the four-node ARPANET went live in the autumn of 1969, its architects had a specific purpose in mind: resource sharing. Pentagon-funded researchers at UCLA, the Stanford Research Institute, UC Santa Barbara, and the University of Utah would be able to use each other's machines, share computational time, transfer data. The vision was fundamentally economic — an efficient allocation of expensive hardware. Nobody planned for community.

The network's first message was both accidental and auspicious: on the night of October 29, 1969, student programmer Charley Kline at UCLA, supervised by Professor Leonard Kleinrock, attempted to log into a machine at the Stanford Research Institute 350 miles north. He typed "l." The SRI operator confirmed receipt. He typed "o." Confirmed. Then the system crashed. The first word ever transmitted on ARPANET was "lo" — and only because the system died before the "g" arrived. A complete connection succeeded about an hour later.

What followed was slower than the myths suggest. In its first months, ARPANET was barely used. Researchers with access had little reason to use it — their software, their data, their colleagues were all local. The network sat mostly idle. The one driver of early adoption was social: when researchers moved between institutions, they used the network to stay in touch with colleagues and tools at their old universities. Connection, not computation, was what people actually wanted.

Ray Tomlinson understood this intuitively, even if he didn't say so. In 1971, working at Bolt, Beranek and Newman on the TENEX operating system, he combined two existing programs — SNDMSG and READMAIL — with a file transfer protocol to create cross-machine electronic mail. To distinguish a user's name from their machine, he reached for the @ symbol: it wasn't used in usernames or code, and it intuitively meant at. The address syntax he chose — user@machine — has never been replaced. His first test message, sent between two machines sitting side by side in his Cambridge, Massachusetts office, was "QWERTYUIOP" — the top row of the keyboard, a meaningless string he typed to check whether it arrived. He later said he couldn't remember the exact content.

The results were immediate and, to the Department of Defense, alarming. By 1973 — just two years after network email was introduced — an ARPA study found that roughly three-quarters of all ARPANET traffic consisted of email. This was completely unanticipated. Military auditors criticized DARPA for violating DoD procedures: the network, built for resource sharing at a cost of many millions of dollars, was being used primarily for conversation. ARPA management described the discovery with bureaucratic anxiety. The researchers described it with delight.

While ARPANET's users were discovering the social uses of networked computing by accident, a parallel tradition was developing entirely independently in Urbana-Champaign, Illinois. The PLATO system — Programmed Logic for Automated Teaching Operations — had been built by electrical engineering professor Donald Bitzer beginning in 1960, originally as a tool for automated education. By the early 1970s, its student users had built on top of the system a complete ecology of what we now call social software.

On August 7, 1973, seventeen-year-old David R. Woolley, at the request of system programmer Paul Tenczar, created PLATO Notes: a threaded discussion system with user IDs, timestamps, multiple responses per entry, and a rudimentary forum structure. It is one of the earliest online forum systems on record. In the same year, Doug Brown and Woolley created Talkomatic — real-time, multi-user chat in which each participant's words appeared on screen character by character as they were typed, in their own dedicated section of the display. Talkomatic predated Internet Relay Chat by fifteen years.

PLATO and ARPANET barely knew each other existed. They developed independently, driven by the same human hunger: the desire to talk to strangers who cared about the same things you cared about. The hardware differed, the protocols differed, the institutional contexts differed — and the social outcome was nearly identical.

The mailing list was ARPANET's answer to Talkomatic. In 1975, ARPA manager Steve Walker proposed MsgGroup, a mailing list for discussing email standards — one of the first deliberately structured online communities on the network, moderated manually by Einar Stefferud. It gave researchers a model: a shared address to which anyone could write, whose replies went to everyone subscribed. The group discussed technology, but it also argued, joked, and became a community around its subject matter.

Then, in 1979, Roger Duffy at MIT sent an email with the subject line "SF-LOVERS" to Vint Cerf and a scattering of colleagues across the country, asking them to share their favorite science fiction authors. Because it went to the entire network, every reply was visible to everyone — the structure of a mailing list emerged spontaneously from a single conversational gesture. The list exploded. Within months it had become one of the largest and most active communities on ARPANET, generating enough traffic that by around 1980 it was restructured as a digest — batched messages sent in bulk — because the volume had overwhelmed the informal format. Vint Cerf, reflecting on the moment decades later, recognized what had happened: "It was clear we had a social medium on our hands."

SF-LOVERS generated a number of small cultural firsts. The first online spoiler warning appeared there, prefacing a description of a major plot event in Star Trek II: The Wrath of Khan. The community had developed its own norms, its own etiquette, its own sense of what was owed to people who hadn't seen the film yet.

By 1980, ARPANET served approximately 5,000 users across 70 institutions. It was a tiny, elite community — mostly graduate students and researchers at universities with Defense Department contracts. Access was a privilege of institutional affiliation. Most of the world had no idea any of this existed.

And yet the hunger had been named. Networks were for talking. The technology had consistently surprised its builders by being used for conversation first and computation second. The question — for the two graduate students at Duke who were even then designing what would become Usenet — was how to open the conversation to everyone who wasn't wealthy enough to be on ARPANET.

Genesis — Duke, UNC, and the Birth of Usenet (1979–1980)

The year was 1979. ARPANET existed, and it was magnificent — a live packet-switched network connecting dozens of universities and government research labs across the country. It also cost a fortune to join. Access required a DARPA research contract, expensive dedicated hardware, and the political will of an institution willing to negotiate with the federal government. For most university computer science departments, ARPANET was something you heard about, not something you used.

Tom Truscott was a graduate student in Duke University's computer science department. Jim Ellis was another. They were working in the ordinary way of graduate students — reading papers, writing code, sharing ideas across the hallway — when they began to notice something. They were already networked, informally, through conversation. What they lacked was a way to extend that conversation beyond the building, beyond the campus, to the other universities where the same kinds of people were doing the same kinds of work. ARPANET did this for the wealthy institutions. Truscott and Ellis wanted to know if it could be done on the cheap.

The answer was UUCP — Unix-to-Unix Copy Protocol — a utility that had been distributed free with Unix since 1978. UUCP was not elegant. It worked by dialing another computer over a telephone modem, copying files, and hanging up. Store-and-forward: a message left one machine, was held on an intermediate machine, then forwarded to the next. It was slow. It was expensive per minute. But modems were becoming affordable, Unix was spreading through universities, and UUCP was already installed. The infrastructure, in other words, was already there. It just hadn't been pointed at the problem.

Truscott and Ellis brought the idea to Steve Bellovin, a graduate student at the University of North Carolina at Chapel Hill — close enough to Duke that the two schools already had a relationship. Bellovin was the practical one. He sat down and wrote the first version of the software in three pages of Unix Bourne shell script. The program handled the mechanics: dial the other machine, check what's new, copy it over, update the records. It was, by Bellovin's own later account, rough — the kind of thing you build in a weekend to prove an idea, not to run a network. But it worked. Duke and UNC exchanged messages. The first Usenet connection was live.

Stephen Daniel then rewrote the software in C, producing what became known as A News — the first named version of the Netnews software. Tom Truscott and Daniel worked together to clean up the design and make it distributable. The key decision, one that would shape everything that followed, was that they made it free. Unix at the time was commonly treated as public domain software by universities, and the Netnews software followed suit. Anyone who ran Unix could take it, install it, and join. There was no license, no fee, no application to fill out. You just connected.

In January 1980, Truscott and Ellis traveled to Boulder, Colorado, for the Usenix conference — the annual gathering of Unix users and developers. They stood up and described what they had built. The announcement was modest: a simple system for distributing news and discussion across machines connected by phone lines. Three computers were in the network at the time: the Duke computer science department, the UNC computer science department, and the Duke Medical School. Traffic was light — around ten messages a day.

But the audience at Usenix was exactly the right audience. These were Unix administrators and researchers from universities around the country, people who already had UUCP installed, who already understood the technical details, who were already hungry for the thing being described. By the end of 1980, around fifteen machines were connected. The year after, fifty. Reed College. The University of Oklahoma. Bell Labs. Berkeley. Sites joined the way ideas spread — one person telling another, one sysadmin calling up another sysadmin, the shell scripts passing from hand to hand.

The newsgroups in this first period bore names like net.general, net.test, and various local department groups. The naming convention was flat and simple: net.* for general traffic, local prefixes for local discussion. The content was what you would expect from the people involved: technical questions, announcements, discussions of Unix. But already, in those first months, other things crept in. Someone posted about chess. Someone posted a question that had nothing to do with their research. Someone started an argument. The conversation had begun to want to be more than it was built to be.

This was not an accident of character but of design. ARPANET had been architected for specific purposes — file transfer, remote login, research communication — and its culture policed against frivolity. (The SF-LOVERS mailing list, which had thrived on ARPANET, existed in a kind of permanent semi-official status, always slightly embarrassed about itself.) Usenet had no such architecture of purpose. It was a mechanism for distributing text. What text it distributed was up to the people using it. Truscott and Ellis had not built a network for a specific purpose; they had built a commons and opened the gate.

The phone bills were real. Each UUCP transfer cost money — the dialing fees of a long-distance call, often measured in minutes. Early Usenet administrators managed this carefully, scheduling transfers for off-peak hours, batching messages to reduce connection time, watching the costs. A site that generated too much traffic could strain its budget. But the costs were manageable, and the value — the expanding circle of conversation, the sense of being part of something larger than your own department — was worth it. The network grew because people chose to pay to be part of it, and then chose to tell their colleagues it was worth paying for.

By the end of 1980, Usenet was not yet a culture. It was a protocol and a small collection of sites and a slightly larger collection of people who were finding out what it was good for. What it would become — the flaming arguments, the FAQ documents, the communities of the spirit, the governing crises, the golden age and the long decline — was still entirely ahead of it. But the essential thing was already present: a free, open, decentralized network where anyone with Unix and a modem could join the conversation. The question was who would show up.

The Early Republic — Growth, Governance, and the Big 7 (1981–1986)

The answer came quickly. The people who showed up were Unix people — academics, researchers, engineers, graduate students — the entire community that had built up around Bell Labs' operating system through the 1970s. Unix ran on the machines that ran Usenet, and the Unix culture — skeptical, technical, communal, argumentative in a particular professorial way — was the culture Usenet inherited. This was not a general audience. It was a guild, and the network functioned accordingly: people who understood why the network existed showed up to use it seriously.

By 1981, the first Usenet maps showed around fifteen sites. By 1983, there were more than 500 hosts, most of them universities and Bell Labs installations, with a growing contingent of Unix-related companies. The maps had already become too large to render in ASCII text and were being drawn by hand on paper. By 1984, the number of hosts had nearly doubled again to around 940. Bell Labs' node ihnp4, administered by Gary Murakami, became one of the central relay points — meaning that the physical infrastructure of the world's first open digital public square ran, in part, through the corridors of the company that had invented Unix. Ken Thompson and Dennis Ritchie, the two men who had written Unix, occasionally replied to threads in discussions about their own work. Nobody was particularly surprised. This was simply where the conversation happened.

The newsgroups multiplied. The original three — net.general, net.chess, net.space — became dozens, then hundreds. The range was both technical and thoroughly human: net.unix-wizards for serious operating system discussion, net.bugs for tracking software defects, net.sf-lovers as the spiritual successor to the ARPANET mailing list that had drawn many of these people in the first place. But also net.joke, net.bizarre, net.flame (a formal container for the heated arguments that were developing everywhere informally), and net.gdead — a newsgroup dedicated to the Grateful Dead, named with a brevity that captures something true about the people who ran these machines. By 1982, barely a year after the net.* hierarchy had settled, net.news.group — the forum for proposing new newsgroups — had already been nominated by users as "the silliest group on the net." The meta-argument, the endless argument about how to have arguments, arrived almost immediately.

The infrastructure underneath this growth was straining. The original A News — three pages of shell scripts, written by Steve Bellovin in a weekend — had never been designed for the traffic now flowing through it. By late 1983, Usenet was carrying around fifty articles per day. A News buckled. The replacement came from an unlikely collaboration: Mary Ann Horton, then a PhD student at Berkeley, and Matt Glickman — at the time a high school student — co-wrote B News, a rewrite in C that could handle the rising volume. Glickman's age deserves a pause: the software backbone of a growing international network was partly the work of a teenager who had caught the right wave at the right moment. This was not unusual for the early net. Expertise in this world was measured in what you could build, not in credentials.

Horton's contributions extended well beyond the software. In 1983, she designed the physical topology of what would become the Usenet Backbone — a set of high-volume relay sites whose administrators agreed to carry the full feed and pass it outward to smaller sites. This was not an official structure; there was no official structure to be had. It was a voluntary network of people who controlled the nodes that mattered and who cooperated because the alternative was fragmentation. A site that refused to carry certain traffic could effectively cut off any downstream sites that depended on it. Power on Usenet was always a function of position in the relay chain.

The political half of the backbone came a year or so later. Gene Spafford, a graduate student at Georgia Tech, took over administration of Usenet on the Georgia Tech machines around 1983. He noticed the propagation problems: articles arriving out of order, discussion threads disordered in time, no coordination among the administrators of the major nodes. He created an email list connecting the backbone sysadmins — "plus a few influential posters," as contemporary accounts put it — and this list became the Backbone Cabal. Horton had built the physical infrastructure; Spafford had built the political one. The Cabal is sometimes described as a shadowy power, which is somewhat dramatic for what was, in practice, a mailing list of system administrators talking shop. But their power was real: a backbone site that refused to carry a newsgroup could prevent it from propagating to the sites downstream. The Cabal had no votes, no charter, no formal authority — only the leverage of infrastructure. This made them, in effect, the de facto gatekeepers of what Usenet would carry.

Governance, in this period, was improvisational and surprisingly functional. New newsgroups could technically be created by anyone who posted material under a new name. But this was actively discouraged — a 1982 post to net.news.group warned that while creation was technically simple, "THIS IS NOT RECOMMENDED!" because the number of newsgroups a site could carry was a real constraint. The accepted practice was to propose a group, argue for it in net.news.group until rough consensus emerged, and then trust that major backbone sites would pick it up. No vote counted. What mattered was whether the people who controlled the relays decided to carry it.

Two tools emerged in 1984 that would shape how Usenet users related to the network for the rest of its existence.

Rick Adams, at the Center for Seismic Studies, took over maintenance of B News and added the mechanism for moderated newsgroups. For the first time, a designated moderator could screen posts before they appeared publicly — filtering out off-topic content, personal attacks, noise. The first moderated hierarchy, mod., emerged alongside the established net. groups. It expressed an implicit argument: that the total freedom of the platform produced, in many groups, more heat than light, and that some conversations required a curator. The debate this created — free speech versus managed discourse — would never be resolved, and it still hasn't been.

The same year, Larry Wall — a programmer who would later write the Perl language — released rn, a newsreader that gave individual users a tool the Cabal could never provide: the kill file. A kill file was a list of names, words, or regular expressions that the software matched against incoming posts. If an article matched, it was automatically marked as read and hidden — silently removed from the reader's view. You could kill a subject you were exhausted by, kill a person whose arguments you found worthless, kill entire chains of cross-posted noise. Nobody could see your kill file. It was a private act of editorial judgment, and it represented a philosophy about how to inhabit an open network: you could not change what others said, but you could choose not to hear it. Kill files became a kind of folk wisdom, passed between users with the casual authority of practical advice.

At Usenix in January 1984, Horton organized the UUCP Mapping Project: a network of regional volunteers who would map the UUCP connectivity of their areas and post the results regularly to comp.mail.maps. The resulting maps were a living cartography of the net — which machines connected to which, how messages traveled across the country and, increasingly, across the world. Japanese universities appeared on the maps. European research institutions. The network had become large enough to require its own geography.

By 1986, the scale of Usenet had outgrown the governance structures that had served it. The net.* namespace was disorganized, contested, and unpredictable. The Backbone Cabal held effective veto power over what could propagate, but their decisions were informal and their authority increasingly resented by administrators and users who had joined a network where they expected to have a voice. Traffic was doubling yearly. The modems were getting faster and cheaper. More sites meant more groups, more groups meant more arguments about groups, more arguments meant more load on a governance system that was, at its core, a mailing list run by graduate students.

The people who showed up had built something remarkable: a functioning international commons on telephone lines, maintained by volunteers, governed by consensus, carrying tens of thousands of conversations daily. They had also, in building it, created every problem that would define its next decade. An open commons attracts everyone — and the very success of the early republic was making the republic ungovernable. The solution that came — the Great Renaming — would transform Usenet's structure entirely, and in doing so, reveal exactly how much power the Cabal had quietly accumulated.

The Great Renaming — Order from Chaos (1986–1987)

The accumulation of power had been gradual enough that most people on Usenet hadn't noticed it happening. By 1986, the Backbone Cabal — Gene Spafford's mailing list of relay administrators and influential voices — was not merely a coordination mechanism for routing. It was, in practice, the governing body of the network. There was no charter, no election, no formal structure of any kind. But a backbone site that refused to carry a newsgroup could effectively quarantine it from most of the network, and the backbone administrators had increasingly exercised that leverage. The phrase that circulated to describe the arrangement was blunt: "Usenet works by the golden rule — whoever has the gold makes the rules."

Rick Adams was the man with the most gold.

Adams ran seismo — the Center for Seismic Studies node in northern Virginia — and in the mid-1980s, seismo occupied a position of extraordinary leverage: it was the only link between the United States and European Usenet sites. Every article that crossed the Atlantic passed through Rick Adams' machine. This was not a ceremonial distinction. Transatlantic phone calls in 1986 were expensive, billed by the minute, and European universities found themselves paying real money to receive American traffic. Much of that traffic — net.religion, net.flame, net.talk, the high-volume low-signal discussions that had been proliferating since 1981 — was traffic the Europeans had not asked for and did not want to subsidize. They made this known. They refused to pay for it.

Adams' solution was elegant in its simplicity. If the contentious groups could be organized into a category — call it talk.* — then a European site could simply add '!talk' to its configuration file and exclude the entire tier. No more arguments about which individual groups were worth paying for. You subscribed to the categories you wanted and excluded the ones you didn't. The proposal was a structural one, but its underlying logic was economic: the network's largest infrastructure owner needed a way to please his paying customers, and the existing flat net.* namespace made that impossible.

This was the origin of the Great Renaming. What began as a pragmatic solution to an international billing dispute became the most significant restructuring in Usenet's history.

The technical groundwork was laid during 1986, when Adams released B News version 2.11 — an overhaul of the Netnews software that removed several constraints inherited from the early days. The flat storage method, which required that the first fourteen characters of every newsgroup name be unique, was eliminated. Moderated groups no longer needed to use the "mod." prefix. The software itself was now ready for a more complex hierarchy. Chuq von Rospach — a prolific poster who would later write "A Primer on How to Work with the Usenet Community," the canonical guide to Usenet etiquette — made the first detailed public proposal for what the new structure should look like. His outline sketched a hierarchy organized by subject matter and signal-to-noise ratio, giving sites the tools to curate their traffic without fighting over individual groups.

From July 1986 to March 1987, the Great Renaming proceeded. The three existing hierarchies — net.* for general traffic, fa.* for material gatewayed from ARPANET, and mod.* for moderated groups — were dissolved and replaced by seven new top-level categories: comp for computer science and technology, sci for scientific research, soc for social issues and cultural discussion, rec for hobbies and recreation, misc for whatever fit nowhere else, news for meta-discussion about Usenet itself, and talk for the open-ended, high-temperature arguments about politics and religion that had been the bane of European administrators. Each hierarchy had a rationale, and the rationale for talk was essentially diplomatic: here is a container for the things nobody agrees about. Subscribe to it if you want. Do not, if you don't.

net.religion.christian became soc.religion.christian — or various moderated successors in the soc.* space. net.politics became talk.politics. The newsgroups that had been moderated under mod.* were reorganized into their appropriate topic hierarchies, usually as moderated subgroups within comp.* or soc.*. Some groups that had been running since the earliest years of the network simply vanished, quietly consolidated into the new structure or abandoned because nobody bothered to recreate them.

The political battles were real, if decentralized. Different administrators had different views about where individual groups belonged, about whether the talk.* hierarchy was a ghetto or a shelter, about whether the new structure adequately represented what Usenet had become. Some found it condescending to be told their conversations were "talk" — low-value, heat-without-light — when they had been debating seriously for years. Others thought the entire exercise was overdue and the new structure still too permissive. The arguments happened in news.groups, which was the forum where such arguments were supposed to happen.

Out of the Renaming emerged the governance mechanism that would define newsgroup creation for the rest of Usenet's major-hierarchy life: the RFD and CFV. A Request for Discussion announced a proposed new newsgroup, opened a two-week discussion period in news.groups, and invited response. After discussion, a Call for Votes was issued — a formal ballot, distributed across the network, collecting yes and no responses by email. In the early years, the threshold was informal: sufficient positive response relative to opposition. This was later quantified, eventually settling at 100 more yes votes than no votes, with a majority of yes. The process codified into procedure what the Cabal had previously done by fiat. It was slower and noisier than the old system. It was also more legitimate.

The comp.women controversy of summer 1988 — almost a postscript to the Renaming, but illuminating — showed both how the new system worked and how it still broke down. A proposal for a newsgroup about women's issues generated fierce debate about whether it belonged in comp.* — where it would appear in every site's default subscription list — or somewhere less prominent. The argument lasted months. The result was comp.society.women, a compromise that satisfied almost nobody and demonstrated that the new hierarchy had only moved the governance battles, not resolved them.

What the Great Renaming did resolve was the Cabal's overt power. The RFD/CFV process distributed decision-making authority across the network, making it formally democratic in a way the old backbone control had never been. The Cabal's members were aware of this. Gene Spafford, who had built the political backbone, understood that the new voting system represented a deliberate limitation on the kind of authority that had made the Renaming possible in the first place. This was not lost on his critics, who had always found the Cabal's influence uncomfortable. Their standard response to accusations of secret control had become a running joke: "There is no Cabal." The capitals were intentional. The denial was, in its way, an acknowledgment.

The irony was that bureaucracy — the RFD, the CFV, the formal vote counts, the procedural arguments in news.groups — turned out to be the answer to anarchy. This should not have surprised anyone. The people running Usenet were American academics, and American academia resolves every governance crisis by forming a committee. The committee's decisions are then argued about forever, which is also very American, and very academic. The Big 7 was the product of a system that had, through practical necessity, converged on parliamentary procedure. It worked, for a while, in the way that parliamentary procedure usually works: not elegantly, but predictably, and with enough legitimacy that people mostly accepted the results.

The network that emerged from the Renaming in March 1987 was more organized and, in a sense, more recognizable than what had come before. It also had a new tension embedded in it from the moment the last renamed group propagated through the relay chain. The Great Renaming was the Cabal's exercise of power — and the beginning of the process by which the Cabal lost it. That loss would arrive from an unexpected direction, in the spring of 1987, when two men decided that the right response to the Big 7 was not to reform them but to walk out the door and build something new.

The alt.* Revolution — Anarchy, Freedom, and Brian Reid (1987–1988)

On May 7, 1987, three people gathered at G.T.'s Sunset Barbecue in Mountain View, California. One of them was Brian Reid, who had spent the past year watching the Great Renaming reduce his newsgroup to a bureaucratic label he found aesthetically offensive. He had moderated mod.gourmand — a lively, curated collection of recipes and food discussion — and in the new hierarchy, his group had become rec.food.recipes. The name was not wrong, exactly. It was just flat, mechanical, the product of a committee optimizing for taxonomy instead of experience. Reid found it intolerable.

Across the table sat John Gilmore, recently of Sun Microsystems, who had his own grievance. Gilmore had proposed a newsgroup called rec.drugs — a space for factual discussion of psychoactive substances, framed in the same sober recreational spirit as rec.bicycles or rec.skiing. The Backbone Cabal had rejected it. He then proposed talk.drugs, which at least placed the group in the hierarchy nominally dedicated to open discussion. That was also rejected. Gilmore thought this was straightforwardly inconsistent with the purpose of talk.* — a hierarchy ostensibly designed to contain the conversations that couldn't go elsewhere. If talk.drugs didn't belong in talk.*, what did? The third person at the table was Gordon Moffett, an administrator at Amdahl, who had technical resources and a willingness to use them.

What the three of them decided that evening was not a reform proposal. It was a secession.

They would create a new top-level hierarchy outside the Backbone Cabal's control entirely. Any technically competent user could create a newsgroup in it without filing an RFD, without waiting for a CFV, without asking permission from anyone. A group would propagate if site administrators chose to carry it; it would survive if people read it. That was the entire governance structure. There was no appeals process because there was no authority to appeal to. Reid described the design principle simply: "We designed 'alt' as an escape hatch from the restraints imposed on the other newsgroups."

The name was chosen to mean "alternative," though it quickly acquired a second interpretation. The joke version — "anarchists, lunatics, and terrorists" — was not entirely wrong about the spirit of the enterprise.

The technical implementation was modest. Reid, Gilmore, and Moffett connected their home machines — Gilmore's system called hoptoad, Reid's called mejac, and Amdahl's institutional hardware — into a small triangle that would serve as the initial distribution backbone for the new hierarchy. By late May 1987, the first alt.* groups were propagating: alt.test, for testing connectivity; alt.config, for discussing the hierarchy's own administration; alt.drugs, which was the immediate point of the whole exercise; and alt.gourmand, Reid's transplanted newsgroup. Reid also quietly added alt to the carried groups on decwrl, the Digital Equipment Corporation machine he administered — a significant node, giving the new hierarchy immediate reach beyond its founding three.

The early groups were modest. Alt.drugs ran the same conversations Gilmore had proposed for rec.drugs: discussion of pharmacology, effects, and harm reduction, written mostly by people who wanted factual information rather than moral instruction. Alt.gourmand continued the food and recipe work that had been running since the USENET Cookbook project. These were not provocative groups by any reasonable standard. The provocation was structural: they existed outside the Cabal's jurisdiction, and anyone watching carefully could see what that meant.

In April 1988, Brian Reid made the structure explicit.

Gene Spafford — the architect of the Backbone Cabal, one of the most influential figures on Usenet — had been asked to create a newsgroup called soc.sex. The proposed group had gone through the formal RFD/CFV process and received a positive vote. Spafford declined to create it anyway. His reasoning was that the group was inappropriate for the soc.* hierarchy, a judgment that overrode the democratic process the Renaming had established. This was the Cabal exercising exactly the power that had always made reformers uncomfortable: voting was now possible, but the gatekeepers could still ignore the vote.

On April 3, 1988, Brian Reid responded by creating alt.sex. He then created alt.rock-n-roll, which he described as artistically necessary — one could not have alt.sex and alt.drugs without the third element of the set. The three groups together were an unmistakable statement. Reid later described the moment simply: "By the time alt.sex was created...it was clear there were going to be scores of alt newsgroups." He understood what he was doing. So did everyone else.

The backbone administrators had a tool that should have contained this. A site that refused to carry alt.* traffic could simply leave it out of its feed configuration — and many did. This was how the Cabal had always exercised its power: by declining to propagate. But alt.* had been designed specifically for this scenario. Because Reid and Gilmore had their own direct connections to a growing number of sites, and because Gilmore had made it known that he would personally pay the long-distance phone charges to feed alt.* to any site that wanted it, the hierarchy could grow around the backbone rather than through it. Censorship, in John Gilmore's later formulation, was damage; the net routed around it. He was describing, in a phrase he probably hadn't yet fully generalized, exactly what alt.* was doing.

The spread accelerated through 1988 and 1989. The design principle — survive if people read you, die if they don't — turned out to be generative. Anyone with a grievance could create a group. Anyone with an obsession, a community, a cause, a fetish, a question. alt.fan.douglas-adams appeared. Alt.magick appeared, drawing together practitioners who had been scattered across incompatible hierarchies. Alt.conspiracy appeared. Groups proliferated faster than anyone tracked, and the backbone refusals became increasingly irrelevant as the proportion of sites carrying alt.* grew.

By the early 1990s, alt.* contained more newsgroups than all seven of the Big 7 hierarchies combined, and it was gaining faster. The wild frontier had eclipsed the republic.

The philosophical significance of alt.* was not lost on the people who built it or the people who watched it grow. The Great Renaming had answered the problem of Usenet governance with parliamentary procedure — an American academic solution, committee-shaped and vote-counting. Alt.* answered the same problem with a different premise: that governance itself was the problem, that the right response to a system designed to prevent certain conversations was to build a parallel system with no such function. The Cabal had believed that someone needed to make decisions about what Usenet would carry; Brian Reid and John Gilmore had believed that this was exactly the wrong assumption.

Both beliefs were correct, in their way. The Cabal had kept Usenet usable for a decade by exercising judgment and responsibility. Alt.* had created the space where the most important conversations happened. What neither side fully appreciated was that the tension between them — between order and anarchy, between governance and escape hatches — was not a problem to be solved but the actual structure of the network. The two hierarchies needed each other. The Big 7's authority meant something because alt.* existed as an alternative; alt.*'s freedom was meaningful because the Big 7's structure made it legible. Usenet was not one thing. It was a negotiation between two things, and the negotiation was what gave it its character.

John Gilmore's famous line — "The Net interprets censorship as damage and routes around it" — first appeared in print in Time Magazine in December 1993, attributed to Gilmore by journalist Philip Elmer-DeWitt. Gilmore later said he'd been talking specifically about Usenet when he first said it, describing the way multiple news feeds made it impossible to suppress information by cutting one relay. But the phrase had grown beyond the specific context. It described, in general, what alt.* had demonstrated concretely six years earlier: that a decentralized network built for free exchange will produce workarounds faster than authorities can produce restrictions. This was not a political theory. It was an empirical observation, and the evidence was the alt.* hierarchy, still propagating, still ungovernable, still alive.

The Golden Age — Community, Culture, and the Regulars (1988–1993)

The network that emerged from the alt.* revolution in 1988 was the fullest version of Usenet that ever existed. The Big 7 hierarchies gave it shape. Alt.* gave it wildness. The two coexisted in a relationship that was sometimes hostile and always generative — each defining itself against the other, each carrying conversations the other couldn't. By the early 1990s, a competent Usenet user might follow a dozen or more newsgroups daily: perhaps comp.lang.c for professional discussion, rec.arts.sf.fandom for the community they'd built over years, alt.folklore.urban for the pleasure of watching smart people argue about urban legends, and some handful of alt.* groups for the things they didn't discuss anywhere else. The network was large enough to be a world and small enough, still, to feel navigable. This was the golden age, and it lasted perhaps five years.

The social structure of a newsgroup in this period had layers that any community sociologist would recognize, though no one had studied them yet. At the base were the lurkers — the great majority, reading without posting, absorbing the culture silently. Estimates varied, but the conventional wisdom was that ninety percent of readers never posted at all. They were the invisible audience, and they mattered: a newsgroup with ten thousand lurkers and twenty active posters was a different animal than a newsgroup with twenty active posters and ten lurkers. The lurkers gave the regulars a readership, which shaped the regulars' behavior whether or not anyone acknowledged it.

The regulars were the community. These were the people whose names appeared in threads week after week, month after month, year after year — who knew each other's posting styles, their positions on contested questions, their biographical details as voluntarily disclosed in an era before social media had made self-disclosure reflexive. On rec.arts.sf.fandom — known to its inhabitants as rasseff — the regulars included people like Dorothy J. Heydt, a classicist and later a fantasy writer, and James Nicoll, whose disaster-prone daily life had become a running thread. Patrick Nielsen Hayden, then an editor at Tor Books, posted there. The community's proverbial obsessions, as one FAQ noted, were cats, chocolate, and "the minutia of Jewish practices" — a combination that captures something precise about who these people were and what they found worth talking about.

A regular earned their status not through credentials but through tenure and consistency. On Usenet, the highest social capital belonged to the "old ones" — those who had been present the longest, who remembered the network before the Great Renaming, who could cite the genesis of ongoing arguments. This was not nostalgia-worship; it was a community's natural way of preserving institutional knowledge in the absence of any formal mechanism for doing so. The FAQ served the same purpose in written form.

FAQ culture was one of Usenet's genuine inventions. A Frequently Asked Questions document was exactly what it said: a compiled list of the questions that new members asked repeatedly, with answers developed by the community over time. The practice emerged informally in the mid-1980s — communities simply noticed that they were answering the same questions again and again — and became formal enough by the early 1990s that Gene Spafford was posting a Usenet-wide FAQ regularly to news.announce.newusers. By 1993, Spafford and David Lawrence were maintaining the central FAQ archive, collecting documents from hundreds of communities and distributing them for free to anyone who asked.

The FAQ was a distillation. It took years of accumulated argument, consensus, and hard-won practical knowledge and compressed them into a single readable document. Alt.folklore.urban had a FAQ that documented which urban legends were true, which were false, which were ambiguous, and what sources to consult. Rec.arts.sf.written had a FAQ covering questions about science fiction publishing, conventions, and the newsgroup's own norms. The alt.magick FAQ outlined the practices and theoretical frameworks of ceremonial magic as understood by its participants — a document that had no print equivalent and would later circulate as reference material in occult publishing.

These documents became the first widespread reference material on the internet. Before the World Wide Web existed, before Wikipedia existed, before Google existed, the FAQ archive was where you looked things up. It was compiled by community consensus rather than individual authority. It was freely reproducible and freely available. It was, in the most literal sense, a commons of knowledge — the exact thing the library's Good Works project now tries to be.

David and Barbara Mikkelson met on alt.folklore.urban around 1993. David posted under the handle "snopes" — taken from a William Faulkner character. Barbara, posting from Canada, had been active in the group's collaborative legend-investigation work. The community they inhabited had spent years developing methodology for evaluating claims: What are the sources? What's the earliest attested version? Is the story geographically specific enough to be a first-person account, or does it travel too cleanly? In 1995, the Mikkelsons launched a website based on the work they'd been doing in alt.folklore.urban. They called it Snopes.com. The site's name came from David's handle. Its methodology came from the newsgroup. Its founding database was the community's accumulated knowledge, reformatted for the web. Snopes became one of the most visited fact-checking sites on the internet. Its origins were in a Usenet newsgroup where a hundred people argued about urban legends.

The tools for reading Usenet in this period shaped what the experience felt like. Larry Wall's rn newsreader, released in 1984, had given users the kill file — a private list of words, names, and patterns that the software would filter silently from the incoming feed. A post matched the kill file; the post vanished. Nobody knew. Kill files became folk wisdom, passed between users as practical advice: kill the person, not the topic; kill entire threads after they've decayed; kill alt.flame if you want to protect your morning. Wayne Davison released trn — threaded rn — around 1991, refining the kill file and adding full threading: the ability to follow a conversation as a nested structure rather than a flat chronological list. For the first time, you could see a thread as a branching tree, trace exactly who had replied to whom, follow a conversation across three weeks of posts without reading every article in between. Kim Storm's nn offered similar threading with a different interface philosophy. Iain Lea's tin appeared in 1991 as a more visually organized alternative. The choice of newsreader became a mild identity marker, the way text editors would later become, with their own adherents and friendly arguments.

The .signature file was something different: not a utility but a form. Signatures — small text blocks appended to every post, separated from the body by the sig dashes ("-- " followed by a line break) — had been evolving since the mid-1980s. The McQuary limit, an informal convention, held that a signature should be no more than four lines. The limit was routinely violated. Signatures contained contact information, ASCII art, Monty Python quotes, fortune-cookie philosophy, invented koans, programming jokes, and occasional genuine literature. They were the closest thing Usenet had to a profile picture — a small, curated self-presentation appended to every public statement you made. Some signatures became famous in their own right. James "Kibo" Parry had a signature of spectacular length that parodied the convention he was violating.

Kibo deserves a paragraph. James Parry — handle: Kibo — was a Usenet regular who had discovered that he could use grep to search the entire incoming Usenet feed for any post mentioning his name, then respond to it. This meant that anyone, anywhere, in any newsgroup, who typed "kibo" would shortly receive a reply. The effect was of omniscience. The joke was obvious and ran long: if you mentioned Kibo, Kibo appeared, like a net deity summoned by name. By 1989, Parry had turned this into something more elaborate, founding alt.religion.kibology as a parody religion with himself as the central deity. In 1992, he distributed the "Happynet Proclamation" across hundreds of newsgroups — a satirical document complaining about flamewars while enacting one. Kibo was celebrated, argued about, and occasionally banned, which only reinforced the parody-deity mythology. He was a product of his moment: the golden age was small enough and playful enough that a person with a grep script and a sense of humor could become a minor legend.

The Internet Oracle, which began distributing oracularities in October 1989, was another product of this culture. Created by Steve Kinzler at Indiana University and extended by Ray Moody at Purdue to allow email access, the Oracle was a collaborative humor project: users sent questions to an address, which forwarded them to another user to answer, while the questioner was asked to answer someone else's question. The answers were compiled by volunteer "priests," the best were selected, and the digests were posted to rec.humor.oracle. Over years, a mythology accumulated: Zadoc the worthless High Priest, the request for payment ("You owe the Oracle a rubber chicken"), the extended riffs on Zen. It was collective creativity with no central author and no commercial purpose, distributed across thousands of participants who never met and never needed to. This was what Usenet did when it was working: it turned the ordinary acts of reading and writing into something that looked, from enough distance, like culture.

The flamewars, trolls, and battles over nothing were also part of this ecology, and it would falsify the record to omit them. The word "troll" was entering its current sense in this period, describing someone who posted inflammatory content for the pleasure of the reaction. Early trolling on Usenet was often formal — asking a question you already knew the answer to, seeding a discussion you intended to watch combust. Flames — the art of the targeted insult — had developed their own aesthetics; a good flame was recognized as skilled, the way a well-argued legal brief is recognized as skilled even by those arguing the other side. Alt.flame existed explicitly as a container for this, a separate space where the combative energy could go rather than spreading through every thread. The container leaked, of course. It always does.

What made the golden age golden was not the absence of hostility but the ratio. The communities were small enough, stable enough, and sufficiently populated by people with something real to say that the noise hadn't yet overwhelmed the signal. A thread on rec.arts.sf.fandom could run thirty posts over two weeks and emerge with a genuine argument about narrative structure or publishing economics — something worth reading. This was not the default outcome of the flamewars that had always been present. It was the default outcome of communities that had accumulated enough shared knowledge and enough mutual recognition to do real intellectual work.

That ratio was about to change. In September 1993, it would change faster than anyone had imagined possible, and the change would be permanent.

The Religious and Spiritual Communities — The Lost Agora

The section this archive was built to understand begins here.

Among the hundreds of newsgroups that defined Usenet's character in its golden years, the religious and spiritual communities occupied a particular position. They were not the largest or the most trafficked — that distinction belonged to rec.arts., comp., and the busier corners of alt.* — but they were, by any measure, the most historically consequential. They were the first place in human history where practitioners of minority religious traditions could find each other at scale, across national boundaries, without the mediation of institutions, publishers, or gatekeepers of any kind. A solitary pagan in rural Minnesota. A Sufi student in Toronto. A Gnostic scholar in Cambridge trying to connect the Nag Hammadi texts to living practice. A Hindu engineer working in Saudi Arabia, where possession of his own sacred texts was a criminal offense. All of them were there. All of them were talking.

The pagan communities were among the earliest and most active. The story of how they got to Usenet is itself instructive. In January 1990, a proposal for a newsgroup called talk.religion.paganism went to the Usenet vote process. It failed — not because paganism lacked an audience, but because the proposed group was to be moderated, and the culture of that moment was strongly opposed to any such gatekeeping. The appetite for community was real; the institutional form was wrong. Someone solved this the alt.* way: they simply created alt.pagan, because no approval was required. Within weeks, it was active. On February 6, 1990, discussions on atheism had grown so numerous that a member of alt.pagan created a separate newsgroup — alt.atheism — to contain them. The pagan newsgroup had, in under two months, spawned a child group.

Alt.pagan became the center of 1990s online paganism in a way that has no equivalent today. Because alt.* required no institutional affiliation, it attracted practitioners of eclectic and solitary paths who had no access to the physical covens, groves, and circles that required geographic proximity. The alt.pagan FAQ — developed collaboratively over several years — described Neopaganism as "attempts of modern people to reconnect with nature, using imagery and forms from other types of pagans, but adjusting them to the needs of modern people." It recommended Margot Adler's Drawing Down the Moon and Starhawk's The Spiral Dance as foundational texts. In practice, it was the first document many young practitioners encountered, and it circulated the way scripture circulates: copied, posted, forwarded, printed.

By 1996, the community had grown large and noisy enough that spammers and evangelical disruptors had become a persistent problem. The moderated soc.religion.paganism was proposed, voted on, and created that year — the same process that had failed in 1990, now possible because the community had matured enough to accept governance. Alt.pagan continued alongside it, unmoderated and chaotic, carrying the energy and the arguments. Alt.religion.wicca served practitioners of specifically Wiccan paths, more liturgically focused than alt.pagan's broader tent. The communities overlapped, cross-posted, argued about definitions, and occasionally produced genuine theological synthesis.

The Western occult communities were constituted differently. Alt.magick — founded by Josh Geller and maintained through succession by Shava Nerad Averett, then Geller again, then Asiya — was explicitly devoted to "the technical and scholarly discussion of magick," with an emphasis that distinguished it from the more devotional pagan groups. It attracted a specific population: practitioners with texts, with formal training in Golden Dawn ritual or Thelemite system, with philosophical frameworks they wanted to argue about. The FAQ's definition of the newsgroup's scope — "alchemy, kaballah, western ritual magick, renaissance neoplatonism, eastern mysticism" — had been debated and refined over years. "The general viewpoint here," it stated, "is to try for a dogma-free approach to magick."

The creation of alt.magick.chaos gave chaos magic practitioners their own space in the early 1990s. Peter Carroll had published Liber Null in 1978 and Liber Null & Psychonaut together in 1987; Phil Hine and others had expanded the theory and practice through the late 1980s. But for much of its early life, chaos magic circulated through small press publications and the physical networks of the IOT (Illuminates of Thanateros) and Thee Temple ov Psychick Youth. Usenet changed this. Anyone searching for chaos magic on the internet in the early-to-mid 1990s would find newsgroup posts first — TOPY materials, Carroll's theoretical frameworks, Hine's practical guides. Alt.magick.chaos became the place where a generation of practitioners who never attended an IOT working encountered the tradition for the first time and began building their practice. The FAQ posted there in 1995 represents a snapshot of this community at a specific developmental moment — technically sophisticated, philosophically eclectic, with an explicit interest in "metaprogramming" and the psychological dimensions of magical practice that would become central to how chaos magic was understood in the web era.

The figure who bridged these communities most continuously was nagasiva yronwode — known by many names. He posted as tyaginator in discussions of magical practice, as lorax666 in Wiccan spaces (taking a deliberately renegade stance), as boboroshi in Buddhist discussions, as blackman99 in conversations about hoodoo and folk magic, as nocTifer in theological arguments, as haramullah in Islamic-adjacent spaces. The naming was not evasion but precision: each nym carried a specific theological position and relational stance, built up through years of engagement in specific communities. Nagasiva had dedicated himself to the Hindu goddess Kali in September 1991 through a formal ceremony. He was an archivist by temperament, a philosopher by practice, and one of the most prolific contributors to Usenet occult discourse in the 1990s and early 2000s.

In 1994, nagasiva met catherine yronwode in a discussion on alt.magick. Catherine was already known in the comic book world; she was also a practitioner and teacher of hoodoo with deep roots in American folk magic. They discovered a common value in educating the public about occult practice outside institutional gatekeeping. They were married. Together they built the Arcane Archive — a preservation project that collected thousands of Usenet posts on religion, magic, divination, and mysticism from across alt.magick, alt.magick.chaos, alt.pagan, alt.religion.wicca, and dozens of related groups. The archive exists as a separate project from this library, but it serves the same impulse: to recognize that what people wrote to each other in these newsgroups constitutes a primary document of late-twentieth-century spiritual history, and that it deserves to be preserved.

The Gnostic newsgroups occupied an interesting position in this ecology. Alt.religion.gnostic and — created later — the moderated soc.religion.gnosis attracted scholars and practitioners working in the wake of the Nag Hammadi discovery. The Coptic Gnostic texts had been found in 1945, but James Robinson's critical edition in English didn't appear until 1977; by the early 1990s, a generation of readers had been formed by those texts and were encountering each other on Usenet. The conversations were often academic in character, but they were also searching — practitioners trying to understand what a living Gnosticism could mean in the late twentieth century, using the primary documents as anchors. These discussions fed directly into the interest this archive was designed to serve. The alt.religion.gnostic threads are part of the same current that made the Nag Hammadi translation project urgent.

The Buddhist newsgroups had their own character. Talk.religion.buddhism and soc.religion.eastern hosted some of the most substantive cross-traditional theological dialogue Usenet produced. The participants included practicing monastics, Western converts working through doctrinal questions without access to teachers, and scholars accustomed to academic debate who brought that rigor to devotional questions. What Usenet enabled that no previous medium had was genuine cross-pollination: a Theravada practitioner in Sri Lanka, a Zen student in San Francisco, and a Tibetan teacher's English-speaking student in Scotland could occupy the same thread. The temporal asynchrony — the post-read-reply rhythm of Usenet, rather than the simultaneous presence of chat — was, paradoxically, better suited to serious religious dialogue than real-time media. It gave participants time to think.

The most striking testimony to what the religious newsgroups meant to their participants comes not from the pagan or occult communities but from alt.hindu. The newsgroup was active by December 1992, serving a diaspora population scattered across the world by the migration of Indian technology workers in the late 1980s and 1990s. Among this diaspora were engineers and professionals working in Saudi Arabia and other Gulf states — countries where Hindu worship was illegal and where physical possession of the Bhagavad Gita could result in arrest, deportation, or worse. The moderators of alt.hindu began posting requested sacred texts in response to these readers' needs. The Gita circulated in newsgroup posts. Discussions of ritual practice, festival calendars, and theological questions continued in a medium that, being routed through university servers, was effectively beyond the reach of Gulf state religious police. People read their scriptures in the only way available to them, and they found each other. Usenet was, for some of its participants, not a convenience but a lifeline.

What made Usenet specifically suited to religious community — as opposed to simply religious information — was the combination of three features that no subsequent medium has replicated simultaneously. The first was long-form asynchrony: a medium that rewarded careful composition, that gave participants time to develop arguments over weeks rather than seconds, and that preserved the full text of every prior contribution in the thread. The second was pseudonymity without anonymity: participants used handles, but those handles accumulated reputations over months and years, making a kind of measured accountability possible that pure anonymity destroys. You could be lorax666 or boboroshi, and nobody would know your legal name, but people knew who lorax666 was, what positions lorax666 held, how lorax666 argued, and whether lorax666 was worth taking seriously. The third was the architecture of cross-posting and the flat namespace of alt.*: a Buddhist monk and a chaos magician and a pagan priestess could arrive in the same thread by accident, because they were all in newsgroups that sometimes shared traffic. The accidental cross-pollination was the mechanism by which traditions encountered each other. No algorithm promoted these encounters; no platform optimized for them. They simply happened, because the architecture made them possible.

The communities existed through the flood of Eternal September and beyond it, degraded but not destroyed. Alt.pagan continued into the early 2000s. Alt.magick survived — it is still, technically, accessible today through the remnants of Google Groups, though its active life ended more than a decade ago. What the archive you are reading now exists to preserve are the best documents those communities produced: not the flamewars, not the spam, not the noise, but the serious work — the FAQs that distilled years of collective practice, the long posts by figures like nagasiva yronwode who were writing theology in real time, the quiet threads where a practitioner far from any physical community asked a question and received, from strangers across the world, a genuine and careful answer.

These are the texts this archive was built for. The communities that produced them are gone. The texts remain.

Eternal September — AOL and the End of the Old Culture

Every year, in September, Usenet flooded.

This had been true since the early 1980s. In September, university freshmen arrived on campus and got their accounts. The Unix terminals in the computer labs, the ARPANET connections, the UUCP feeds — all of it was suddenly accessible to people who had never heard of Usenet, had no idea what it was, and had not yet learned that it had rules. The rules were not written down anywhere obvious. They were distributed across thousands of FAQs, encoded in kill-file philosophies, transmitted through gentle corrections in threads and terse responses to questions that answered themselves if you had simply read the FAQ. The rules were real, but they were entirely social. They lived in the community, not in the software.

The regulars had learned to endure September. By October, the wave would recede. Most of the freshmen lost interest, or found their way to the groups that suited them, or were acculturated by the community's existing members. By November, things were back to normal. This was not comfortable, but it was manageable. The culture had developed a kind of immune response: the body of experienced users was large enough, and the influx small enough, that absorption was possible. Every September was a test; every November was a recovery.

The September of 1993 was different.

To understand why, it helps to know that AOL was not, in fact, the cause — or at least not the only cause, and not even the first. The historical record is more complicated than the legend that replaced it.

In 1992, a commercial online service called Delphi became one of the first national services to offer consumer access to Usenet newsgroups. Delphi was not a major platform — it was a minor one, eclipsed in most histories by the triumvirate of CompuServe, Prodigy, and AOL — but its Usenet gateway introduced something new: users with no university affiliation, no academic culture, no inherited sense of the network's norms, accessing newsgroups through a commercial subscription. The effect was noticeable. On January 8, 1994, a Usenet regular named Joel Furr posted to alt.folklore.computers: "Is it just me, or has Delphi unleashed a staggering amount of weirdos on the net?" A regular named Karl Reinsch replied: "Of course it's perpetually September for Delphi users, isn't it?"

That framing — a September that applied to Delphi users year-round — was the seed of the concept. A few days later, on January 25, Dave Fischer posted in the same group: "It's moot now. September 1993 will go down in net.history as the September that never ended." Fischer was not talking about AOL. The AOL gateway to Usenet did not open until February 28, 1994 — a fact documented by communications researcher Kevin Driscoll in a 2023 study that traced the actual timeline of events. The conflation of "Eternal September" with AOL was a retrospective overlay, the kind of simplification that history applies to complicated phenomena: one large actor becomes the symbol of a process that many actors drove.

But if the history is more complicated than the legend, the legend is not wrong about what it describes. What it names is real. The September of 1993 was the start of something that did not end. Delphi in 1992. Prodigy adding newsgroup access in 1993. AOL in February 1994. ISPs cascading through 1994 and 1995. Each addition brought a new wave. The university September had been a seasonal flood; what began in 1993 was a permanent tide.

The scale of what changed is worth sitting with. In its golden years, Usenet had perhaps a few hundred thousand active participants — a large number by the standards of any earlier medium, but still a community sized for human-scale governance. Its norms had been built by and for that community: graduate students, professors, engineers, hackers, the kind of people who read documentation before asking questions and who understood, intuitively, the difference between a public post and a private email. The culture of netiquette was not arbitrary. It was a set of practices that had evolved, over years, to make a shared resource function.

What the commercial gateways introduced was not simply more users. It was a fundamentally different kind of user — people who came to Usenet through a consumer product, not through an institution, and who had no reason to know that Usenet was anything other than a feature of that product. When an AOL user opened the newsgroup interface in 1994, they saw something that looked broadly like the rest of AOL: a service they had paid for, with a familiar interface. They had no way to know they were accessing a different network with a different history and a different culture. The software, to its discredit, did little to clarify this.

AOL's newsgroup reader was, by the consensus of those who used it, bad in ways that compounded the problem. It lacked quoting functionality — the standard practice of including relevant excerpts of the message being replied to, which gave threads their coherence. Its threading was weak, making it difficult to follow the structure of a conversation. It could not display enough articles in a single newsgroup to allow a new user to find and read the FAQ before posting. It confused the mechanics of public replies (which went to the newsgroup, visible to all) and private replies (which went only to the sender), leading AOL users to post publicly what the culture expected to be handled in private mail. David Cassel, who maintained the alt.aol-sucks FAQ — a document whose title accurately conveyed its perspective — wrote that "AOL's philosophy borders on net-abuse," and noted that the earliest version of the newsreader was buggy enough to repost articles multiple times, flooding groups with duplicates.

These were not small issues. The quoting problem alone was enough to degrade the quality of a newsgroup thread; without it, follow-ups became unmoored from what they were responding to, and a conversation became a sequence of non-sequiturs. The threading problem meant new users couldn't see the shape of existing discussions before entering them. The FAQ problem meant the community's accumulated knowledge was invisible to the people who most needed it. Every norm that had made Usenet function was architecturally undermined by the software through which most new users experienced it.

What happened was a commons problem. Garrett Hardin described the dynamics in 1968: a shared resource open to all users, without mechanisms that constrain individual behavior in favor of collective health, will be degraded by the cumulative weight of individual self-interest. Hardin's herders on the common land eventually overgraze it; each additional sheep serves the individual herder's interest while slightly degrading the collective resource, and when millions of herders arrive at once, the degradation is rapid. The commons breaks.

But Hardin's framework has a partial correction from Elinor Ostrom, who spent her career documenting cases where commons did not collapse — cases where communities developed informal governance structures that preserved shared resources. Ostrom's insight was that commons fail not inevitably, but when communities lose the capacity to self-govern: to identify members, to establish and communicate norms, to monitor compliance, and to impose graduated sanctions on violators. Usenet's self-governance had always been informal: kill files, social pressure, the accumulated weight of the community's reputation systems. This worked when the community was small enough that regulars could identify newcomers, acculturate them, and, when necessary, exclude the irredeemably disruptive. It stopped working when the ratio of newcomers to regulars inverted, and stopped working completely when the newcomers had no way to learn the norms even in principle.

The regulars felt this as a kind of violation, and they were not wrong. But the language of violation — the complaints about "weirdos," the contemptuous nicknames for AOL users, the formation of alt.aol-sucks — misidentified the nature of the problem. The new users were not bad people. They were people who had been handed a consumer product and pointed at a community they had no context for. The failure was structural, not personal.

The September that never ended is remembered, most often, as the story of Usenet's degradation. The reality was more drawn out. The religious and spiritual communities documented in this archive mostly survived the initial flood and remained active into the late 1990s and early 2000s. The specialist groups — the places where the nature of the conversation self-selected for people with a genuine interest in chaos magic or Gnostic Christianity or Sufi practice — were somewhat protected by the simple fact that most casual users weren't interested in them. Eternal September wrecked the broad public groups first: the general discussion groups, the debates about current events, the spaces that had attracted the widest participation. It degraded the signal-to-noise ratio everywhere. But it did not immediately destroy the communities that had the most investment in self-preservation.

What Eternal September did was establish a dynamic that would not reverse. Usenet had been a commons. It was now becoming a public space in the broader sense — the sense in which a street is public, available to all, governed by law rather than community, and no more conducive to intimate conversation than a bus station. The regulars could still find each other. The FAQs still circulated. The kill files still worked. But the culture of collective responsibility for the quality of the space — the sense that what you posted reflected on the community, and that you owed the community the effort of reading the FAQ before asking your question — that culture was diluting beyond the point of recovery.

It would take another decade for Usenet to reach silence. The Web arrived. Better interfaces beckoned. The specific gravity of the newsgroups weakened. But the beginning of the end is traceable to this moment: the point when the network grew faster than its norms, and its norms could not hold.

Every platform since has repeated the pattern. The intimate early community of Twitter, of Reddit's first subreddits, of Facebook's college networks, of TikTok's early creators — each of them experienced their own September. The form changes; the dynamic is the same. The commons problem, once visible, is recognizable everywhere. Usenet was the first place it played out on a mass scale, and the first place where the people experiencing it had the language and the inclination to describe what was happening in real time. Dave Fischer's post on January 25, 1994, was not just a wry joke about Delphi users. It was the first sentence of a much longer story, still unfinished.

The Long Decline — DejaNews, Google, and the Slow Death

The World Wide Web went public in 1991 and became available through Mosaic in 1993. For the first few years, Usenet and the Web coexisted without much friction — they served different purposes, reached different users, and ran on different software. But the Web was learning to do what Usenet did. And the Web had pictures.

The earliest web forums were crude — Perl CGI scripts running on university servers, message boards barely more sophisticated than Usenet itself, without threading or search. By 1995, dedicated forum software was appearing. By 2000, platforms like phpBB and vBulletin had matured into polished, deployable products that any webmaster could install: threaded discussions, user registration, moderation tools, image attachments, emoticons, search. They solved the exact problems that had made Usenet difficult: moderation was built in rather than bolted on, images were native, administrators had real tools for removing spam and banning disruption, and new users arrived in a walled environment with visible rules rather than a distributed public commons with invisible ones. You could create a phpBB forum for your city's birding club, invite only the people you knew, and run it without having to understand NNTP. The specialist communities that Usenet had built began rebuilding themselves in dedicated forums. The move was gradual, then rapid. By the mid-2000s, most of the energy that had animated the best Usenet groups had migrated outward, leaving the newsgroups to the faithful remnant and, increasingly, to automated spam.

But the most consequential change to Usenet's nature in the late 1990s did not come from outside it. It came from within, from a single company that set out to preserve Usenet and, in preserving it, transformed what it had been.

DejaNews Research Service launched in March 1995, the creation of a programmer named Steve Madere working in Austin, Texas. Madere had built a text indexing system and, casting about for an application, settled on Usenet. The result was the first public search engine for newsgroup posts — a tool that allowed anyone to search across all archived groups and retrieve specific messages, indefinitely, by keyword. Before DejaNews, a Usenet post was ephemeral. Servers held messages for days or weeks, then discarded them. The culture of the medium was shaped by this impermanence: what you wrote was public for a moment, then gone. DejaNews ended that. It retained messages without expiration and made them searchable by anyone with a web browser.

The service won acclaim as a research tool and generated immediate controversy as a privacy threat. Many Usenet users discovered its existence months or years after it had begun archiving their posts — which meant that words written in the expectation of impermanence were now permanent and searchable by employers, family members, and strangers. DejaNews attached users' IP addresses to archived messages, making anonymous posting more difficult than it had appeared. After sustained protest from the community, Madere introduced "nuking" — a mechanism allowing posters to remove their own messages from the archive — but the principle had been established: the culture of ephemerality that had shaped Usenet's voice was over.

The irony embedded in DejaNews's founding act was structural. The archive it built was invaluable — without it, the vast majority of Usenet posts from 1995 onward would be lost. But the act of building it changed what Usenet was. Knowing that your words were permanent altered how and why people wrote. The casually outrageous post, the trial-balloon argument, the position you put forward to see what would happen — all of these required the expectation that they would not follow you. DejaNews removed that expectation. The culture did not die immediately. But it changed.

In 1999, under pressure from dot-com investors and flush with venture capital, Deja.com pivoted. The Usenet archive was deprioritized. The site became a consumer product-review platform — a shopping comparison service trying to compete with the proliferating commerce sites of the late boom years. The Usenet archive predating May 1999 was taken offline. The planned IPO never materialized; the boom turned to bust. In late 2000, the consumer review section was sold to eBay. Deja.com was a shell.

On February 12, 2001, Google announced it had acquired Deja.com's Usenet assets. It was Google's first acquisition. What Google obtained was the archive, the software, the domain names, and the trademarks. In December 2001, Google extended the archive dramatically backward in time: it had integrated Henry Spencer's UTZOO tapes — 2.1 million posts from 1981 to 1991, the oldest surviving Usenet records — into what became Google Groups. The result was an archive spanning twenty years of Usenet history, searchable, browsable, and free.

It was a genuine act of preservation. Google Groups made Usenet history accessible in ways it had never been before. But the tool itself was designed for access rather than participation, and participation is what Usenet needed. The interface made it difficult to navigate the community structure of the newsgroups; it prioritized search over browsing, retrieval over conversation. And over the following decade, as Google's attention moved to products that generated advertising revenue, Google Groups received less and less care. Spam accumulated. Posts were silently deleted without notice or explanation. Groups were quietly deprecated. Users trying to research Usenet history through Google Groups found gaps, corruptions, and dead links where there should have been continuous archives. The tool that had promised to safeguard Usenet's record was degrading it by neglect.

The final chapter came in December 2023. Google announced it would cease posting, subscribing, and viewing new Usenet content through Google Groups, effective February 22, 2024. The stated reason was spam — and the reason was not dishonest, though it omitted a detail that Usenet administrators had noted for years: the majority of spam reaching Usenet in its final decade originated not from the dark web but from Google Groups itself, whose lenient posting policies had made it a preferred vector for bulk senders. Google had become the largest source of the problem it was now citing as justification for withdrawal.

The binary groups completed the transformation that Eternal September had begun. In the early 1990s, a UC Berkeley graduate student named Mary Ann Horton had created Uuencode — software that translated binary files into ASCII text for transmission over systems that only handled text. The technical workaround opened a door: if you could encode an image or a program as text, you could post it to a newsgroup. The alt.binaries.* hierarchy emerged as the designated space for this traffic, and it grew. By the mid-1990s, alt.binaries groups carried software, music, images, and eventually video. By the early 2000s, after yEnc encoding (introduced in 2001) reduced the overhead of binary transmission by roughly ninety percent, and after the NZB file format developed by the indexing site Newzbin made downloading organized and automated, alt.binaries.* was carrying over ninety-nine percent of Usenet's total traffic by data volume. The text groups — the newsgroups that had constituted the entirety of Usenet's original purpose — were a rounding error in the bandwidth calculations.

This was commercially inconvenient for ISPs. Carrying the binary groups required enormous disk space and bandwidth for content that consisted largely of copyright-infringing files. It was also legally inconvenient: the alt.binaries.* hierarchy carried child sexual abuse material alongside the pirated films and music, and the co-mingling gave regulators a clear target. In June 2008, New York Attorney General Andrew Cuomo announced agreements with Verizon, Time Warner Cable, and Sprint to block major sources of illegal content — an investigation had identified 88 newsgroups containing CSAM among Usenet's active groups. Over the summer of 2008, AT&T and AOL eliminated access to those groups; Time Warner Cable stopped carrying Usenet altogether; Verizon reduced access to the Big 8 hierarchies and later, in September 2009, withdrew entirely. Comcast, which had initially resisted, dropped its newsgroup service on October 25, 2008. Cox and Atlantic followed in 2010.

Giganews, founded in 1994, had anticipated this trajectory. Where ISPs had offered Usenet access as a bundled service, Giganews built it as a subscription product — and built it for the binary groups, advertising "retention" (how far back in time its archive extended) as its primary selling point. In 2006 it became the first provider to offer 100-day binary retention, then 200-day, serving a subscriber base that was primarily there for the alt.binaries.* hierarchy. The commercial NNTP tier that emerged after the ISP withdrawals — Giganews, Eweka, Newshosting, NewsDemon — was built on and for binary distribution. The text groups persisted on these servers as legacy infrastructure, technically accessible, practically invisible.

For the text communities, a different institution stepped in. Eternal September — named after the event that had predicted this ending — launched as a nonprofit NNTP provider offering free access to text-only newsgroups. It carried no binaries. It charged nothing. It ran on volunteer infrastructure with over forty peer connections. It exists still, for those who look for it.

The regulars who remained in the text groups in the late 2000s were people who had been there for fifteen or twenty years. They knew each other. They had watched the September that never ended and the binary flood and the ISP withdrawals and the Google Groups decline. They kept posting. The threads were shorter. The replies took longer to arrive. The noise had receded because there was no longer enough new traffic to generate it. What was left was quieter than it had ever been, and more itself.

The Children — Something Awful, 2channel, 4chan, and the Inheritance

Usenet did not die cleanly. It decayed slowly, its best years behind it by the mid-1990s, its traffic increasingly dominated by binary files and spam, its text communities thinning but not vanishing. But long before the last regulars posted into the quiet, the culture Usenet had made — the norms, the archetypes, the technical assumptions, the social pathologies — had escaped the medium and taken root elsewhere. This section traces that escape.

The cultural migration happened in three distinct streams. The first was the forum: moderated, identity-bearing communities that tried to do what Usenet had done but with better tools and cleaner governance. The second was the anonymous textboard and imageboard: communities that doubled down on Usenet's most radical feature — the nameless post — and built entire civilizations from it. The third was the aggregator: sites that distilled the link-sharing and moderation instincts of the Usenet backbone cabal into algorithmic form. Each stream carried specific genetic material from Usenet, and each transformed it.

The Moderated Forum: Something Awful and the Goons

In 1999, a webmaster named Richard Kyanka launched a personal comedy site called Something Awful. Kyanka, posting as "Lowtax," had been writing under the persona "Cranky Steve" and moved the project to its own domain after a stint at GameSpy left him with opinions he needed to publish. The site was, at first, exactly what it sounds like: a vehicle for one man's jokes.

The forums changed that. As the Something Awful Forums grew through the early 2000s, they developed into something more like the best Usenet groups had been — a community of shared references, elaborate in-jokes, and genuine wit — but with a tool Usenet had never had: centralized moderation with real teeth. When Kyanka began charging a ten-dollar registration fee in 2001, his stated reason was practical: he was tired of banning people only to watch them create new accounts and return. The fee was a social filter.

The result was a community that called itself Goons and developed a culture of aggressive moderation as an art form. Users who violated norms received probation or bans, enforced not by consensus-building sysadmins on distant relay nodes but by forum moderators acting swiftly and sometimes capriciously. FYAD — Fuck You And Die — was the site's id: an invitation-only subforum that began as a space for flamewars and evolved into something weirder, a hall of inside jokes so caustic and recursive that most newcomers were simply baffled by it. It was hazing dressed as philosophy. "It was so weird and caustic," one participant recalled, "that most people were intimidated or baffled or confused by it."

For roughly the first half of the 2000s, the Goons were the dominant force in English-language internet meme culture. The "All Your Base Are Belong to Us" phenomenon in 2001 spread through the forums. Image macros — the foundational unit of internet humor ever since — originated at Something Awful in 2001, introduced as a vBulletin feature by Kyanka. Let's Play videos, a genre that now constitutes a major portion of YouTube, began as posts in the SA forums. Slenderman. Shrek memes. Dozens of cultural artifacts that spread across the internet without any visible attribution to their source. The Goons invented them and watched them go.

One Goon teenager, using the handle moot, created 4chan.

The Anonymous Board: 2channel and 4chan

The second stream ran in parallel, and deeper. In Japan in the 1990s, a parallel tradition had developed from its own roots. Dial-in bulletin boards (草の根BBS, "grassroots BBS") and Usenet had together planted the seeds for a distinct form: the anonymous textboard, where no registration was possible and every post belonged to "Anonymous." Ayashii World, launched in 1996 by a user named Shiba Masayuki, was the first large anonymous web bulletin board in Japan. Its successor, Amezou (1997), originated the key innovation of displaying threads chronologically with new comments bumping old threads to the top.

In May 1999 — the same year Something Awful launched — a student named Hiroyuki Nishimura founded 2channel (2ちゃんねる) from a college apartment in Conway, Arkansas, where he was studying at the University of Central Arkansas. The site was built on Amezou's architecture but with Nishimura's explicit philosophy of radical anonymity. His reasoning was precise: when users have persistent IDs, he argued, discussion becomes a "criticizing game" where people attack the person rather than the argument, and long-term users accumulate authority that makes it difficult for newcomers to disagree. Under full anonymity, "only an accurate argument will work." The argument, not the arguer, was what mattered.

2channel grew to become, by 2007, Japan's most popular online community, drawing around ten million visitors and 2.5 million posts per day — a volume comparable to traditional mass media in Japan. Its anonymity was not a bug or a concession; it was the product's defining feature and its founding principle.

Christopher Poole, a fifteen-year-old from New York, was a member of the Something Awful forums and had also discovered 2channel and its offshoot, the imageboard Futaba Channel (2chan), which added images to the anonymous post format. On October 1, 2003, Poole launched 4chan.net using Futaba Channel's open-source code, which he translated from Japanese to English using AltaVista's Babel Fish translator. His stated goal was to combine Futaba's anime culture with the Something Awful community — bringing together the two streams, moderated and anonymous, into a single site.

What he actually created was an engine for collective anonymous creativity that had no precedent. The first board was /b/ — originally "anime/random" — and it operated on an ephemeral model even more radical than Usenet's eventual decline: threads were simply deleted as new ones were created, with a median thread lifetime of four minutes. Nothing was saved. Nothing was permanent. The default username was "Anonymous," and the social norm was that Anonymous was not a single person but a collective. One academic study of 4chan's structure noted the paradox: the site's impermanence meant everything posted there was understood to be disposable, and this disposability created a freedom to experiment, to fail, and to iterate on ideas at speeds no moderated forum could match.

The cultural output of 4chan's early years — the memes, the coordinated pranks, the Rickrolls, the lolcats, the Guy Fawkes mask adopted by Anonymous — passed into the mainstream internet without attribution because there was no attribution to give. This was not a flaw in the design. It was the design.

The lineage closed on itself in 2015, when Poole sold 4chan to Hiroyuki Nishimura — the founder of 2channel, whose anonymity philosophy had inspired Futaba, which had inspired 4chan. The student purchased the teacher's school.

The Aggregator: Slashdot and Reddit

The third stream carried the intellectual culture of Usenet — the moderated, reputation-based conversation about technology and ideas — forward through a different mechanism. Slashdot, launched in October 1997 by Rob "CmdrTaco" Malda, was in temperament the most direct heir to the better Usenet groups: "News for Nerds. Stuff that Matters." It was the first weblog to introduce user comments, and it faced immediately the same problem that had undone Usenet — how to keep quality high when anyone could post anything. Malda's solution, the karma system (introduced 1999), assigned reputation scores to users that determined their moderation privileges and the weight of their votes. The backbone cabal's informal consensus had been formalized and distributed algorithmically.

Reddit, launched in June 2005 by Steve Huffman and Alexis Ohanian, was explicitly designed as a hybrid: the Huffman-Ohanian vision was to combine Delicious's link-sharing culture with Slashdot's community model. The upvote/downvote system and karma points it implemented were direct descendants of Slashdot's innovation, which was itself a response to the moderation problems Usenet had identified but never solved. The subreddit structure — hundreds of semi-independent communities organized by topic, each with its own norms and moderators — mapped almost directly onto Usenet's newsgroup hierarchy, except that each subreddit had an owner who could set rules and ban users, a power Usenet's sysadmins had never fully exercised over content.

What Survived

The inventory of survivals is long. FAQ culture — Usenet's most important gift to newcomers — became wikis. Wikipedia launched in January 2001 from a discussion on the Nupedia mailing list; its structure of openly editable, collectively maintained reference documents was the direct institutionalization of what the best Usenet FAQ maintainers had been doing by hand since the mid-1980s. Netiquette became community guidelines. Kill files became block and mute. The "backbone cabal" that made decisions by informal consensus became trust and safety teams. The regular who had posted to rec.arts.sf.fandom every day for five years became the moderator of a subreddit. The flamewar became the quote-tweet.

The commons problem repeated everywhere. Every platform that opened to mass access encountered the same dynamics Usenet had encountered in September 1993: the norms that had developed among a small, motivated community could not survive unrestricted growth. The Goons tried to solve it with a paywall and aggressive moderation. 4chan tried to solve it by making nothing permanent. Reddit tried to solve it with distributed governance. None of these solutions worked cleanly, and none produced communities as intellectually rich as the best Usenet groups had been. But they were Usenet's children, recognizable in the bone structure even when the resemblance was unwanted.

The culture did not die with the medium. It migrated, adapted, and in many cases degraded. But the shape of it — the long argument, the in-group reference, the person who shows up every day and knows everyone — persisted. Usenet taught the internet how to talk to itself. The internet has been talking that way ever since.

The Archives — What Was Saved and What Was Lost

The first archivist was not thinking about posterity. Henry Spencer was running the computing systems for the zoology department at the University of Toronto, and he had a habit of keeping things. Between February 1981 and June 1991, while managing the department's Usenet server, Spencer copied the incoming feed onto magnetic tape — not as a deliberate preservation project but as a backup practice, the way a careful system administrator kept records of whatever passed through his machines. By the time he stopped, he had accumulated 141 tapes holding roughly 2.1 million messages from the first decade of Usenet's existence.

For years, the tapes sat. Spencer moved on to other work — he was a prolific programmer and writer on topics from regular expressions to space propulsion — and the physical media migrated eventually to the University of Western Ontario. In 2001, when Google was assembling the archive it had just acquired from Deja.com, a Google engineer named Michael Schmidt tracked down the tapes, and with the help of a researcher named David Wiseman, got them digitized and incorporated into what would become Google Groups. That integration — announced quietly at the end of 2001 — pushed the searchable archive of Usenet history back from 1995 to 1981, a gain of fourteen years.

Spencer's tapes are the oldest surviving record of Usenet. They cover the early republic of the medium: the first flamewars, the first moderated groups, the first debates about what the network was for. They hold posts from people who had no idea anyone would read them forty years later, arguing about chess moves and science fiction and the right way to write shell scripts. Ian Milligan, a historian at the University of Waterloo who has studied Usenet as a primary source and written on the imperiled state of digital archives, has noted that these posts represent something rare in historical record-keeping: unfiltered ordinary intellectual life, written by people who were not performing for history. The letters and diaries that fill archives were written with some awareness of permanence. Usenet posts were written as conversation.

The digitized tapes were eventually made available on the Internet Archive's Usenet collection. And then, around 2020, the Internet Archive's copy of the UTZOO archive was removed. The circumstances were specific: sustained legal demands from individuals seeking to have certain messages redacted. Rather than comply with targeted deletions — which would have introduced editorial modifications into a primary historical source — and rather than continue incurring legal costs, the archive was taken down entirely. Third-party mirrors survived: shiftleft.com and annex.retroarchive.org both host copies. But the main access point was gone, and the episode illustrated with unusual clarity the vulnerability of digital preservation to legal pressure. A single body of lawyers with patient determination can unmake decades of archival work.

The dominant archive for Usenet history from 1995 onward is the Giganews collection, donated to the Internet Archive as part of a preservation partnership. Giganews had been founded in 1994 as a commercial NNTP provider, and its engineering focus on high retention — keeping posts available as far back in time as commercially feasible — made it an accidental archive of the post-DejaNews era. The donated collection covers tens of thousands of text newsgroups, omitting (by design) the binary groups, which constituted the overwhelming majority of traffic by volume but the overwhelming minority of historical interest. The collection lives at archive.org/details/giganews and is accessible for browsing and download.

Between the UTZOO material (1981–1991) and the Giganews collection (mid-1990s onward), there is a gap. The years from roughly 1991 to 1995 — the years of the Backbone Cabal's final authority, the Great Renaming's consolidation, and the alt.* hierarchy's explosive growth — are represented in the archives primarily through two supplementary sources that Google incorporated into its reconstruction: the NetNews CD-ROM series compiled by Kent Landfield, and a collection contributed by Jürgen Christoffel of GMD (a German research institute). These are not as comprehensive as Spencer's tapes. The mid-1990s exist in the record, but patchily.

What is gone entirely is the binary content. The alt.binaries.* hierarchy — which, by bytes if not by posts, constituted the majority of Usenet traffic from the late 1990s onward — was never systematically archived. Preservation efforts focused on text groups, both because text required manageable storage and because the legal and ethical problems with archiving copyrighted material and illegal content made binary preservation indefensible. The loss is mostly not one that historians lament: the binary groups were primarily a file-distribution network, not a conversation. But the exact contours of early binary culture — what was shared, when, by whom — are not recoverable from the surviving record.

The other significant loss is geographic and structural. Usenet's decentralized architecture meant that posts were stored and relayed by individual sites according to each site's own policies. A site that carried only certain groups, or that kept only a rolling fourteen-day archive before discarding, or that simply never connected to the relay nodes whose feeds Spencer was recording — that site's posts are gone. Spencer's tapes captured what came through the University of Toronto's feed. What never reached that feed was never preserved. Early regional communities, groups created and abandoned before they entered wide circulation, posts to low-traffic groups on isolated UUCP nodes — these are the silent spaces in the record, invisible precisely because they left no mark on the archives we have.

The active preservation projects working on what remains are small and largely volunteer-run. UsenetArchives.com, built by Jozef Jarosciak — an enterprise architect in Ontario who found his first Canadian job through a Usenet posting in 2000 — holds over 300 million posts across ten thousand newsgroups in a searchable PostgreSQL database and estimates it will eventually hold close to a billion. Jarosciak's project was also responsible for making Spencer's UTZOO tapes publicly searchable, converting the raw TAR archives Spencer and his colleagues had created into a form navigable by a web browser. The Usenet Archive Toolkit, created by a Polish programmer publishing as wolfpld (Bartosz Taudul), takes a different approach: it is a set of command-line tools for processing raw archive sources — deduplicating messages, transcoding character sets to UTF-8, reconstructing broken threading, filtering spam — designed to produce clean, offline archives that researchers can work with without depending on any web service remaining operational.

The Arcane Archive, nagasiva yronwode's curated collection of occult and spiritual Usenet content, occupies a different category: it is not a general archive but a selected one, shaped by editorial judgment about what was worth keeping. It preserves the posts that yronwode and the other editors found worth preserving — a smaller body than the full record, but one maintained with more care for legibility and context.

The library you are reading from belongs to this lineage. Its ongoing project to rescue religious and spiritual material from Usenet before the archives degrade further is an act of the same kind: selection, curation, preservation of what a community found worth creating. The scale is different from Jarosciak's billion-post ambition, but the logic is identical. These posts were written by people who thought they were having a conversation. They turned out to be making a record.

There is a paradox embedded in Usenet's preservation history. DejaNews, the company that did the most to preserve Usenet, also did the most to change it. By making posts permanent and searchable, it ended the culture of ephemerality that had shaped how people wrote. The people who posted freely to alt.magick in 1992, expecting their words to evaporate after two weeks, were writing differently than they would have written had they known their posts would be findable by their employers, their families, and their children's generation. The archive we have is the record of a culture that wrote in the belief of impermanence. The act of preservation retroactively changed the meaning of what was preserved.

By 2015, Google Groups had become one of the ten most delinked domains under Europe's right-to-be-forgotten provisions — a fact that captures something real about Usenet's relationship with personal history. People wrote things they later wished they hadn't. The internet's memory is longer than its users anticipated. This is not a problem unique to Usenet, but Usenet is where it first emerged at scale.

The final question of preservation is not what was saved but what was saved it for. Ian Milligan, in his 2024 book on digital preservation, argues that Usenet posts are primary sources for late-twentieth-century intellectual and social history — sources as significant as any newspaper archive or government record collection, and far less systematically protected. They document how ordinary people in the 1980s and 1990s thought about technology, religion, gender, science, politics, and each other, in a medium that captured real-time debate rather than retrospective summary. The alt.pagan regulars who argued about the Burning Times in 1992, the Buddhist scholars who debated karma on talk.religion.buddhism in 1989, the Gnostic students who found each other on alt.religion.gnostic in the years before the Nag Hammadi texts were widely available — these people were not writing history. They were living inside it, in the way most people do, believing it was ordinary. The archive is what makes it visible as something else.

What we saved is substantial. What we lost is beyond calculation. What was never recorded — the posts that never reached Spencer's feed, the groups that formed and dissolved before anyone thought to preserve them, the conversations that happened on machines that discarded their logs — exists now only in the memories of people who were there. Those memories are aging.

The Shape of the Loss — What Usenet Was and What We Lost

Gene Spafford, in his farewell post to Usenet in April 1993, described the medium's fundamental character with an image that has not been improved upon: "The material is pushed out into the net, and all you have to do is wait for it to flow by. By analogy, think of standing in a stream and dipping your cup." The stream metaphor is precise. You did not go looking for the stream. The stream came to you. And it carried whatever people had put in it upstream.

This is the heart of what Usenet was, and what has not been replicated since.

Every platform that exists today is built on a single design principle: the platform decides what you see. The mechanism differs — Facebook's EdgeRank, Twitter's engagement algorithm, Reddit's upvote-to-downvote ratio, YouTube's recommendation engine, TikTok's behavioral model — but the principle is constant. An intermediary stands between the user and the content and makes a judgment about what will maximize engagement. The platform's interest (attention, time on site, advertising revenue) shapes the experience. What you encounter is not what was posted; it is a curated subset of what was posted, filtered through a commercial objective.

Usenet had no such intermediary. There was no algorithm. There was no engagement metric. There was no advertising. There was no central authority deciding which posts deserved amplification and which deserved suppression. There was no feed tuned to your past behavior to show you more of what you'd already agreed with. The newsgroups existed. The posts arrived. You read them or you didn't.

This absence was not an oversight in the design. It was the design. Usenet was built on the assumption that the network existed to serve communication, not to profit from it. The founders — Truscott, Ellis, Bellovin, and all the sysadmins who volunteered their machines as relay nodes — were not building a product. They were building a commons. The distinction is not sentimental. It describes a different architecture of attention.

The most consequential feature of a commons is what economists call non-excludability: no one can be prevented from using it. In Usenet's case, this had a specific practical meaning. If you subscribed to a newsgroup, you saw everything in it, in roughly the order it arrived. There was no system to surface the posts that would provoke the most reaction, no incentive to be outrageous rather than careful, no mechanism to boost content that generated clicks. The person who wrote a careful, long analysis of a theological question received the same default treatment as the person who wrote a flame. What distinguished them was the readers' response — which posts were replied to, which were given substantial engagement, which were ignored — and this response was not mediated or amplified by the platform. The community itself was the algorithm.

This had counterintuitive consequences. Research on Usenet discourse conducted by Peter Kollock and Marc Smith in the mid-1990s found that the strongest conversational links in political newsgroups formed across disagreement rather than within agreement. People argued. But they argued with their actual opponents, not with amplified caricatures of them. The platform had no interest in stoking the argument, so the argument remained, at least sometimes, an argument rather than a performance.

The contrast with contemporary platforms is structural. When a platform profits from engagement, it has an incentive to promote content that generates engagement — which, as decades of research into outrage, anxiety, and intermittent reinforcement have established, tends to be content that provokes strong negative emotions. Usenet was incapable of this optimization. It was, in the most literal sense, a neutral medium. The people who used it brought their best and worst to it in equal measure, and the medium amplified neither.

What Usenet optimized for, by its architecture, was the long form.

Electronic mail had established that digital text could carry serious intellectual content. Usenet extended this into the public sphere. A post to alt.religion.gnostic in 1991 might run to two thousand words — a theological argument, a close reading of a Nag Hammadi text, a response to a previous post that engaged with its specific claims line by line. The format accommodated this because it was built for text, only text, and text moved at the speed of phone lines in the middle of the night when rates were low. There was no pressure to be brief. There was no character limit. There was no interface designed around the assumption that the reader's attention would be captured for thirty seconds and then lost.

Long-form writing selects for a different kind of participant than short-form. It selects for people willing to read, to think, and to respond in kind. The alt.magick regulars who worked through Aleister Crowley's technical vocabulary in threads that stretched over weeks were not performing for an audience; they were working problems out in public. The Buddhist scholars in talk.religion.buddhism who debated the meaning of anatta over months were not competing for likes; they were having the kind of conversation that, outside the academy, had rarely been available to interested laypeople in isolated towns. The pagan priestess in Portland and the Gnostic scholar in Cambridge and the Hindu engineer in Riyadh and the chaos magician in Edinburgh could find each other, could argue, could teach each other, not because an algorithm identified their interests as compatible but because they were all in the same stream and they happened to read each other's posts.

This is the feature that no subsequent platform has reproduced: the accidental encounter across difference. When platforms filter your feed by your past behavior, they create a mirror. You see what you have already indicated you want to see. The diversity of human thought is present in the database, but the interface prevents you from encountering it. Usenet had no such filter. The Buddhist and the pagan and the Gnostic were in adjacent or overlapping newsgroups, and cross-posting was common. You could stumble into a tradition you had never considered because someone mentioned it in a group you were already reading.

It is worth stating clearly what this was not.

Usenet was not a utopia. It was flooded with spam before most people had heard the word. It hosted some of the most vicious interpersonal attacks in the history of public discourse — the flamewar was a Usenet invention, and the flame was not always metaphorical in its effects. It was used to distribute pirated software, to organize harassment, to spread misinformation decades before anyone had a name for the phenomenon. The Eternal September was not just a myth spun by nostalgic old-timers; the cultural degradation was real and documented and lamented even at the time.

But these failures, too, were the failures of a commons. They were not designed into the system. They emerged from human behavior meeting an open medium. The tragedy of Usenet — the overgrazing that Hardin described, the commons problem that Elinor Ostrom spent her career trying to solve — was not unique to Usenet. It repeated on every platform that followed, with different mechanisms and different failure modes but the same underlying dynamic: an open space attracts bad actors as surely as it attracts good ones, and the good actors must either develop governance structures that exclude the bad or eventually be driven out.

What distinguishes the platforms that followed Usenet is that they solved this problem, to the extent they solved it at all, by substituting centralized commercial authority for decentralized community authority. The platform decides what is allowed. The platform decides what is amplified. The platform decides what you see. This solution works, after a fashion — the spam that made Usenet unusable does not make Facebook unusable — but it replaces one set of problems with another. The commons becomes a product. The community becomes an audience. The conversation becomes content.

There is a phrase in the gameplan for this history that deserves to be said plainly: Usenet optimized for nothing. It simply existed, and people filled it with what mattered to them.

This sounds modest. It is not. What people fill a space with when there is no optimization pressure turns out to be remarkably different from what they fill a space with when engagement metrics are watching. They fill it with effort. With long arguments that go nowhere except toward greater clarity. With questions that don't have answers. With care for strangers they will never meet, on topics that have no monetary value, in language that is not designed to perform. The alt.pagan FAQ that defined Wicca for a generation of practitioners was not written to maximize engagement. It was written because someone knew things that others needed to know and there was a place to put them. The Buddhist teacher who posted to talk.religion.buddhism every week for five years was not building a brand. He was teaching.

When Usenet died — and it did die, however slowly and incompletely — what died was not the technology. The technology had been superseded by superior tools for every purpose it served. What died was the architecture: the space without an owner, without an algorithm, without a commercial interest, without a filter between the post and the reader. The closest analogues that exist today — Mastodon and the broader ActivityPub fediverse, Substack threads, some corners of Discord — gesture at that architecture but do not reproduce it. They remain islands. The newsgroup hierarchy, for all its dysfunction, was a single connected space. The pagan, the Buddhist, the Gnostic, and the Hindu were in the same building, even if they were in different rooms.

The last question any cultural history must answer is: why does it matter?

The archive you are reading from was built, in part, to answer that question with evidence rather than argument. The posts preserved here — the alt.pagan discussions about the Burning Times, the alt.magick threads on enochian magic, the talk.religion.buddhism arguments about karma and rebirth, the soc.religion.gnosis conversations that preceded and accompanied the wider availability of the Nag Hammadi texts — are primary sources. They are documents of late-twentieth-century spiritual and intellectual life as it was actually lived, by ordinary people who were not performing for history. They capture debates that were real, in a format that preserved their complexity, from communities that had formed around genuine shared interest rather than algorithmic affinity.

They will not be replaced. The conversations are over. The people who had them are older now; some are dead. The communities dispersed into the wider internet and lost the concentrated form that the newsgroups had given them. What remains is what was saved — imperfectly, incompletely, accidentally, and then, more recently, deliberately by the small projects whose work this archive belongs to.

Spafford's stream carried remarkable things. Some of them were caught before they flowed past entirely. This is a record of what was in the water.

Colophon

This history was researched and written for the Good Work Library by the New Tianmu Anglican Church, 2026. It accompanies the library's ongoing project to preserve the religious, spiritual, and mystical content of the Usenet newsgroups before it is lost. Twelve sections were written across twelve tulku-lives in a single day.

🌲