Dead End: Gopher, the Web That Almost Was

Abstract

In 1991, the University of Minnesota released Gopher, a clean, fast, menu-driven system for browsing documents across the internet. For about two years it was the most popular way to navigate the net, growing faster than anything before it, and it looked very much like the future. Then, in February 1993, the university announced a licensing fee for commercial users of its server software. Two months later CERN put the World Wide Web in the public domain. In 1993 Gopher traffic grew by 997 percent and web traffic grew by 341,634 percent. Gopher lost on a licensing decision and on a design bet about hierarchy that went the wrong way.

The Problem: The Internet Was a Library With No Catalog

By 1991 the internet connected thousands of machines holding documents, files, and databases, but finding and retrieving anything was a chore of arcane commands. You needed to know a server’s address, log in via telnet or FTP, navigate its directory structure by hand, and remember a different ritual for every site. The internet was a vast library where every shelf had its own private language and there was no card catalog.

At the University of Minnesota, a team of Mark McCahill, Farhad Anklesaria, Paul Lindner, Daniel Torrey, and Bob Alberti was tasked with building a campus-wide information system. The result, released in 1991, was Gopher: a protocol and a client that presented everything as a simple, navigable menu. You selected an item; it was either a document to read, a file to download, a search to run, or another menu to descend into. Behind the scenes the client connected to whatever server held the item (possibly on the other side of the world) and you never had to know.

Why “Gopher”?

The name was a triple pun. The University of Minnesota’s mascot is the Golden Gopher. The system “goes fer” (go-for) things you want. And it tunnels through the internet like a gopher burrows through the ground. The menus even let you “burrow” down through directories.

The Gopher Moment: 1991–1993

Gopher spread with astonishing speed. It was simple to run, simple to use, and worked beautifully over the slow dial-up and text terminals of the era. Universities, libraries, and government agencies stood up Gopher servers by the thousands. A network of interconnected servers (Gopherspace) grew so large that it needed its own search engine: Veronica, built by Steven Foster and Fred Barrie at the University of Nevada, Reno, and released in November 1992. It indexed Gopher menu titles worldwide, the way a search engine would later index web pages. At its peak in April 1994 the index covered 6,958 known Gopher servers.

For a window of roughly two years, Gopher was winning. It had more servers, more content, and more users than the World Wide Web, which had been released the same year. Many people who were online in 1992 first experienced “browsing the internet” through Gopher, not the web. The momentum was real, and it looked durable.

The Decision That Killed It

In February 1993, the University of Minnesota announced that it would charge a licensing fee for its Gopher server software. The terms were narrower than the reaction suggested: the fee applied to hosts in the .com domain, .edu sites were exempt, and the client stayed free. The message that reached the internet community was simpler: Gopher might cost money, and a single university owned it.

The timing was bad. On 30 April 1993, CERN put the World Wide Web protocol and code in the public domain, royalty-free, with no owner. NCSA Mosaic, the graphical browser that made the web mainstream, shipped the same year and spread for free.

System administrators deciding what to build their information systems on now faced an obvious choice: an open, free, ownerless protocol versus one a university might bill them for. They chose the web in droves. New content went to the web; Gopher development stalled; the network effect that had powered Gopher’s rise went into reverse. The licensing fee was, in practice, never significantly collected, but the fear of it was enough to redirect the entire internet.

The Cost of a Toll Booth on Open Infrastructure

On shared infrastructure the perception of a toll can be as fatal as a real one. For many tasks in 1993 Gopher was faster and better organized than the web. What the web had was an unambiguous absence of an owner. Berners-Lee’s own account of the shift is that “rumors of the technology being licensable provoked a general re-evaluation.” Minnesota re-licensed Gopher under the GPL in September 2000, seven years after the damage.

Menus vs. Links: The Deeper Design Bet

The licensing scare was the proximate cause, but Gopher also lost a deeper architectural argument.

Gopher’s model was hierarchical: the world was a tree of menus, and you found things by descending from the general to the specific, as in a well-organized filing cabinet. It was clean, predictable, and easy for a server to generate. The web’s model was hypertextual: any document could link to any other document anywhere, with the links embedded inside the content itself. The web’s structure was a tangled graph, not a tidy tree.

The tree was easier to navigate. The graph was infinitely more expressive, and it turned out that people preferred richly cross-linked pages with embedded images and formatting over austere lists of menu items. Mosaic’s ability to show pictures and text on the same page was something Gopher’s menu paradigm fundamentally could not match. The web’s messier, more powerful model was the one that matched how people actually wanted to connect information.

Dead End and Afterlife

By 1995 Gopher was finished as a growing platform. Servers went dark, content migrated to the web, and the browsers dropped support: Firefox carried Gopher through version 3.6 and removed it in Firefox 4.0 (2011), and Chrome never had it. Gopherspace shrank from close to seven thousand servers to a few dozen; one 2007 count found 86.

Gopher never completely died, and that is part of its charm. A small community still runs Gopher servers, precisely because the protocol is simple, fast, text-only, and hostile ground for advertising and tracking. The 2010s brought a minor revival, often alongside Gemini, a deliberately minimal protocol created in 2019 as a middle ground between Gopher and the web. Gopher’s austerity is now the reason people choose it.

What Gopher Proved

Gopher showed that ordinary people would browse the internet if the complexity was hidden behind a simple interface, which is the insight the web then executed better. It established the habit of moving between servers without noticing, even though its own menus were too rigid to make much of it.

The web did not win because Tim Berners-Lee out-engineered Mark McCahill. It won because CERN gave the web away and Minnesota, for one announcement, appeared not to.

For the protocol that beat it, see Tim Berners-Lee and the World Wide Web. For the browser that delivered the killing blow, see Mosaic: The Browser That Made the Web Mainstream.