← all stories

Before the Web · Era 2 · Into the Internet

1991–presentlost the race for a mass audience to the web by 1994–1995,…

Gopher

Gopher was a system of text menus for finding and retrieving documents on the internet, built in 1991 by a team of programmers at the University of Minnesota who went around a bureaucratic university committee. For two or three years Gopher was the main way to move around the internet, until in 1993 the university announced licensing fees for some of its users, and a few weeks later CERN did the opposite and put the competing web into the public domain, free and without conditions. The world quickly chose free: by 1994 the web had overtaken Gopher in traffic. The protocol itself did not die, and a niche network of a few hundred servers is still running today.

Founders Mark McCahill · Farhad Anklesaria · Paul Lindner · Bob Alberti · Daniel Torrey · Daniel Johnson
Domains boombox.micro.umn.edu (the original University of Minnesota server, 1991 to the 1990s) · gopher.floodgap.com (the main modern server and the project that maintains it)
pre-webprotocoluniversity-projectlicensing-backlashopen-vs-proprietaryniche-survivor

In brief

Gopher was a system of text menus for finding documents on the internet, released in April 1991 by a small team of programmers at the University of Minnesota who went around their own university committee. For two or three years it was one of the main ways to find documents on the network, until in 1993 the university announced licensing fees and its competitor, CERN, in those same weeks put the web into the public domain free of charge and without conditions. The world chose free. By the spring of 1994 the web had overtaken Gopher in traffic, but the protocol did not die: a niche network of a few hundred servers is still running today.

How it started (the founders)

Mark McCahill ran the Microcomputer Center at the University of Minnesota, a unit that in the late 1980s counted as a schism in the university's IT hierarchy, the mainframe priesthood against a handful of personal-computer partisans. Farhad Anklesaria, a programmer with a doctorate in genetics, joined the center in 1987. By 1991 the university had spent several months designing a campus-wide information system through a committee of more than 20 department heads and had produced zero code; Anklesaria recalled it as a "design-by-committee monstrosity."

When McCahill demanded something, anything, to show at the next meeting, Anklesaria cut the protocol down to a bare minimum on an ordinary Mac, without the elaborate search schemes the committee wanted. Four more people joined the project: Paul Lindner (the "Gopher Dude"), Bob Alberti (earlier one of the creators of the MUD game Scepter of Goth, an ancestor of World of Warcraft), Daniel Torrey, and Daniel Johnson, and all six later became co-authors of RFC 1436. The committee met the protocol with hostility, since Gopher did without a mainframe and handed control to the owner of any PC, and the team was barred from working on it for a time. But after the university failed to outsource its computing, the opponents backed off, and in April 1991 Lindner quietly posted Gopher to FTP with an announcement on a mailing list: "Hey, we've got this thing, come and get it."

Year-by-year timeline

Lesser-known but significant facts

  1. The name Gopher is a triple pun taken from RFC 1436 itself: the burrowing rodent, the mascot of the state of Minnesota, and "gofer," the errand boy. The team picked the name specifically so that the committee could not object to it: "It's the school mascot!"
  2. RFC 1436 was being written in the same month as the licensing scandal. The defensive letter of March 11, 1993 refers directly to a nearly finished "informational RFC," which means that standardizing the protocol and the row over monetizing it were happening in parallel rather than in sequence.
  3. One of Gopher's creators, Bob Alberti, had earlier helped build the MUD game Scepter of Goth, an early ancestor of World of Warcraft; he was also the first to propose charging money for Gopher, before the university administration did, and got bafflement from his boss in return.
  4. Port 70, still occupied by Gopher to this day, was assigned by IANA in 1993, the same year as the licensing scandal and the writing of the RFC.
  5. The Gopher team believed that the technically best protocol of the era was not the web at all. In their account the Austrian Hyper-G was superior to both Gopher and WWW but never took off, and simply being at the right point of the wave mattered more than quality.

Legend vs. the record

The first growth lever

The lever was not the power of the university but the lack of it. The committee banned Gopher, and the team released the protocol around it, through FTP and a short announcement on a mailing list, without a single marketing dollar. What worked next was extreme simplicity: a hierarchy of text menus was clear to anybody, an English major with no technical background included, and a server could be set up on any PC, without a mainframe and without permission from the center. Within a year there were already hundreds of servers, and by April 1994 there were 6,958 installations. The second lever came from outside: in November 1992 the University of Nevada, Reno, with no involvement from the UMN team, released Veronica, a free full-text search across all of Gopherspace that solved the main problem of a growing file system, which was navigation. The combination of a zero barrier to entry and free third-party search produced that 997% growth over 1993, which at the time looked unstoppable.

Parallels today (projects from the catalog)

What a builder can take from this in 2026

  1. If you change the rules of monetization, announce the terms by class of user at once, in a single document, instead of by rumor over months. The final UMN scheme was reasonable (free for the non-commercial sector), but it was delivered drop by drop, and the vacuum of information bred a "hysteria" that killed trust faster than the money did.
  2. In the infrastructure layer (protocols, MCP servers, agent skills, formats like llms.txt) an open license from day one is a strategy rather than generosity. CERN's decision to put the web into the public domain was shaped in part by a ready example of the consequences of the opposite move at Gopher, and the same choice today decides the fate of a new agent standard.
  3. Openness wins not because the technology is better but because the friction is lower. Gopher was no worse than the web at the fork of 1992–93; it lost to freer terms of distribution, and in 2026 that is the same lesson for any wrapper over closed LLM APIs facing an open alternative.
  4. An organization with no money to support the infrastructure it made popular is bound to yield to someone with nothing to lose. Six people in Minnesota had neither the resources nor the mandate to maintain Gopher indefinitely, while for CERN openness cost nothing and was almost free PR.
  5. Do not let an unrelated fire at the parent organization take your team away at a competitive fork. The UMN financial scandal of 1995 pulled the entire Gopher team onto an accounting system exactly when the protocol needed development, and the project died from the inside earlier than traffic statistics alone would have shown.

Discrepancies and what we could not verify

Sources (primary first)

Secondary (context, cross-checking):

The same thing, about today

The same breakdowns, but of projects launching right now: what the product is, where the first users came from, how they charge. The card is free, the full dossier is $5 (the dossier itself is written in Russian).

Browse the dossier catalog →