Gopher protocol

» see gopher.floodgap.com as it lived, on the wayback machine

~ the obit ~

gopher was a protocol for organizing and accessing internet content as nested menus. it was released in april 1991 by mark mccahill and farhad anklesaria at the university of minnesota's microcomputer center. the team was trying to build a campus-wide information system. they ended up building a global one by accident.

i never used gopher. i got online in 1998 and by then gopher was effectively gone. the web had won. but for a window of about two years, 1992 to 1993, gopher was the thing that looked like it might be the future of internet content. in early 1993 the rate of new gopher servers being deployed was higher than the rate of new web servers.

the user interface was simple. you connected to a gopher server. you saw a numbered list of items. each item was either another menu, a text document, a downloadable file, or a search interface. you picked one by typing its number or arrow-keying to it. if you picked a menu, you got the next list. if you picked a document, you got the document. it was clean and it worked over very slow connections.

gopher mostly died because the university of minnesota announced in 1993 that it was going to charge a licensing fee for commercial use of the protocol. the announcement was interpreted as the protocol becoming proprietary. the web, by contrast, was free. tim berners-lee at CERN released HTTP and HTML into the public domain in april 1993. by the end of 1993 the web's growth rate was overtaking gopher's. by 1995 gopher was a footnote.

~ the rap sheet ~

~ how it actually worked ~

the gopher protocol was minimal. a client connected to a server on port 70. the server responded with a menu of items. each item had a type code (0 for text, 1 for submenu, 7 for search, 9 for binary, etc.). the client rendered the menu. the user picked an item. the client connected to whatever server hosted that item and fetched it.

the design enforced a tree structure. every item belonged to a parent menu. there was no equivalent of an inline link inside a document. you could not link from one document to another inside the document body. you went up to the menu, picked a different item. the structure was strict but easy to navigate on text terminals.

search was built into the protocol. one of the item types was "indexed search." you would pick it, your client would prompt for a query, the server would return a menu of matching items. veronica (the gopher search engine) indexed gopher servers worldwide and was one of the first internet-scale search systems.

~ the brief peak ~

1992 to 1993 was gopher's window. universities, libraries, and government agencies were building gopher sites at a rapid pace. the white house had a gopher server. the library of congress had one. NASA had several. the EPA, NOAA, and most major US research universities had gopher information systems. you could browse all of them from a single client.

the experience was actually nice for what computers and connections looked like in 1992. terminal connections were common. graphical interfaces were not standard. a text-based menu system that worked the same on a vt100 terminal as on a windows machine was a real value proposition. mosaic and netscape needed graphical computers and faster modems. gopher worked on anything.

the gopher information network ran on a small set of well-known servers. boombox.micro.umn.edu was the main university of minnesota site, and through veronica you could find your way to most other gopher sites in the world. it felt like a coordinated, well-organized library, which is what its designers were aiming for.

~ how the licensing fight killed it ~

in march 1993 the university of minnesota announced that it intended to charge licensing fees for commercial use of gopher. the announcement was vague about what counted as commercial use. some interpretations meant that any business running a gopher server might owe fees to U Minn.

the announcement chilled adoption. companies that had been planning to deploy gopher servers paused. some switched plans to the web, which CERN had explicitly released for free use. by the time U Minn clarified the licensing terms in 1994 (the fees were narrower than first feared), the momentum had moved to the web.

this is sometimes pointed to as the cautionary example of how a small ambiguous licensing decision can kill a technology that had real momentum. it is also more complicated than that. gopher's hierarchical model was fighting an inline-link model that was genuinely better for how people wanted to read documents. even without the licensing dispute, the web's design choices probably win in the long run. but the licensing dispute made the timeline much faster.

~ what it left ~

the few servers that are still up. floodgap.com runs gopher.floodgap.com, which is the main living gopher site as of 2026. there are maybe a few dozen active gopher servers worldwide, mostly run by hobbyists. you can install a gopher client (lynx supports it, gopher: scheme works in some niche browsers) and browse the surviving network. it has a quiet text-only beauty.

the gemini protocol. gemini was designed in 2019 by solderpunk as a deliberate spiritual successor to gopher. it borrows the menu-style layout and the simplicity, while adding TLS, mime types, and a better text format. gemini has more servers than gopher does in 2026. some people consider it the version of gopher that should have existed.

the cautionary tale. gopher is what people point to when they argue that openness matters. when CERN released the web freely and U Minn tried to license gopher, the choice ran one direction. the lesson got internalized by everyone who designed an internet protocol after that. nobody since then has tried to build an internet-scale protocol with proprietary licensing. that is partly because of what happened to gopher.

@}-->--

« previous  ·  random  ·  next »