{"id":162,"date":"2017-08-31T07:10:54","date_gmt":"2017-08-31T10:10:54","guid":{"rendered":"https:\/\/www.linuxpro.com.br\/2017\/08\/a-historia-do-unix\/"},"modified":"2026-09-23T12:20:35","modified_gmt":"2026-09-23T15:20:35","slug":"a-historia-do-unix","status":"publish","type":"post","link":"https:\/\/www.linuxpro.com.br\/en\/2017\/08\/a-historia-do-unix\/","title":{"rendered":"The history of Unix"},"content":{"rendered":"

\"Mascote<\/p>\n

In July 28 2025, a researcher at the University of Utah was cleaning out a storage room when he found a nine-track magnetic tape whose label indicated UNIX version 4. Until then, only the manual had survived from that version of 1973. In December, the Computer History Museum read the tape, and the system ran again in an emulator, more than fifty years later. It's a good reminder of how big this is: a system written for a minicomputer forgotten in a corner of Bell Labs became the ancestor of Linux, the BSDs, macOS, and Android, and to this day anyone who opens a terminal uses its ideas without realizing it.<\/p>\n

<\/p>\n

Note (2026):<\/strong> text by 2017 reviewed and expanded on 2026, with dates checked against Dennis Ritchie's essay, the original article by 1974, and the official UNIX trademark registration.<\/div>\n

1969: Multics closes the door, a PDP-7 opens another<\/h2>\n

In the late 1960, Bell Labs, AT&T's research center, was involved in Multics<\/strong>, an ambitious time-sharing system developed with MIT and General Electric. The project kept slipping and promised more than it delivered. Between 1968 and 1969, the Labs gradually pulled out, and the departure was complete by 1969. A group of researchers, including Ken Thompson<\/strong>, Dennis Ritchie<\/strong>, Doug McIlroy and Joe Ossanna, was left without the interactive environment they had grown used to working in.<\/p>\n

Thompson found a little-used PDP-7<\/strong> from DEC with a nice graphical monitor. First, he and Ritchie ported to the machine a game Thompson had written, Space Travel<\/em>, which cost about 75 dollars of CPU time per round on the lab's big computer. Then came what mattered: a file system that Thompson, Ritchie, and Rudd Canaday had designed on the blackboard. In August 1969, with his wife and son on vacation in California, Thompson spent about three weeks writing the system kernel, the shell, an editor, and an assembler. In the end, the PDP-7 had an operating system capable of developing itself.<\/p>\n

1970\u20131971: a playful name and the first customer<\/h2>\n

The name came later. According to Ritchie, it was \u201cwell into 1970\u201d that Brian Kernighan<\/strong> suggested Unics<\/em>, a mean-spirited pun on Multics: a \u201cuni\u201d system where the other was \u201cmulti\u201d. The spelling became Unix<\/strong>.<\/p>\n

To get a better machine, the group needed a justification. It came from the patents department, which was looking for a system to prepare patent applications. The group offered a text editing and formatting service in exchange for a PDP-11\/20<\/strong>. It worked: in the second half of 1971, three typists from the department spent their day on Unix, while the researchers tried to work in between. The first UNIX Programmer's Manual<\/em>, by Thompson and Ritchie, dates to November 3, 1971<\/strong>.<\/p>\n

1973: pipes and the rewrite in C<\/h2>\n

A third edition<\/strong>, from February 1973, brought the system call pipe<\/code>. McIlroy had been insisting for years on the idea of connecting programs like hoses, the output of one to the input of another. Overnight, utilities began to be written for this, and the style of small programs that do one thing and combine in the shell was born.<\/p>\n

A fourth edition<\/strong>, from November 1973, was the first one with the kernel rewritten in C<\/strong>, the language Ritchie had been developing from Thompson's B. An operating system written in a high-level language was rare at the time; it was what made Unix portable to other machines. In October of that year, Thompson and Ritchie presented the system at the ACM operating systems symposium, SOSP. It is precisely this fourth edition that was on the tape found in Utah.<\/p>\n

1974: the article that spread Unix<\/h2>\n

In July 1974, the magazine Communications of the ACM<\/em> published \u201cThe UNIX Time-Sharing System\u201d<\/a>. The article is short and to the point. It states that since the version for the PDP-11 started working, in February 1971, about 40 installations<\/strong> had been put into use, and that the system occupied 42 KB<\/strong> on a PDP-11\/45 with 144 KB of memory. It presents the ideas that endured: files as byte sequences, devices accessed as files, tree-structured directories, fork<\/code> and exec<\/code>, and the shell as an ordinary program.<\/p>\n

There was also a decisive legal detail. An antitrust agreement of 1956 prevented AT&T from operating outside the telephone business. It could not sell Unix as a product, but it could license it, and it did so at a token price to universities, with the source code included. A generation of students learned operating systems by reading Unix code.<\/p>\n

1975\u20131979: Berkeley and the seventh edition<\/h2>\n

A sixth edition<\/strong>, from 1975, was the first one widely distributed outside Bell Labs. Thompson spent the academic year of 1975\u201376 as a visiting professor at the University of California at Berkeley, where he had graduated, and brought the system along. At Berkeley, the student Bill Joy<\/strong> put together improvements and utilities, such as a Pascal system and the editor ex<\/code> (the vi<\/code> came with 2BSD), and in March of 1978 the first Berkeley Software Distribution<\/strong>, 1BSD, came out.<\/p>\n

In 1979, the seventh edition<\/strong> consolidated the Unix we know: Bourne shell, awk<\/code>, make<\/code>, sed<\/code>, tar<\/code> and uucp<\/code>. Steve Bourne called it \u201can improvement on all previous and subsequent Unixes.\u201d Around the same time, a version for the VAX called 32-bit, 32V, paved the way for 3BSD and 4BSD.<\/p>\n

1980\u20131984: BSD, TCP\/IP, and System V<\/h2>\n

In 1980, Microsoft announced the XENIX<\/strong>, a version of Unix licensed from AT&T, which would become one of the most installed versions on microcomputers of the time. In 1987, Santa Cruz Operation (SCO) took over the product. Microsoft's trajectory during this period is covered in the Bill Gates story<\/a>.<\/p>\n

At Berkeley, DARPA funding paid for the implementation of the TCP\/IP<\/strong>. The 4.2BSD<\/strong>, as of 1983, brought the network stack and the sockets<\/em> API that the internet still uses today. Bill Joy had already left in 1982 to found the Sun Microsystems<\/strong>, whose SunOS was born from BSD.<\/p>\n

At AT&T, the systems group released System III in 1982 and System V<\/strong> in January 1983, already treated as a commercial product, with support. The breakup of the company, on January 1, 1984, finally cut the ties of the 1956 agreement. The family split into two branches: BSD<\/strong>, from the university, and System V<\/strong>, from AT&T. AIX, from IBM, and HP-UX, from HP, followed System V.<\/p>\n

1984\u20131993: the Unix wars and POSIX<\/h2>\n

Each vendor had its own Unix, and programs didn't move from one to another. In 1984, a group of vendors created the X\/Open<\/strong> to define a common environment. In 1987, AT&T and Sun joined forces to unify the market around System V; alarmed, competitors created the Open Software Foundation (OSF) in 1988. The System V Release 4<\/strong>, announced in 1988 and released in 1989, combined features from System V, BSD, and SunOS, and Sun's Solaris 2 became based on it.<\/p>\n

In the middle of the fight, the contract that still matters today emerged: the POSIX<\/strong> (IEEE 1003.1, of 1988), the specification of the interface that a Unix-like system must offer. Later, the Single UNIX Specification<\/strong> began to define what can use the UNIX trademark.<\/p>\n

In 1992, Unix System Laboratories (USL), a subsidiary of AT&T, sued BSDi and the University of California, alleging that Net\/2, the free BSD distribution, contained AT&T code. The lawsuit froze free BSD for almost two years. In June 1993, Novell bought USL, and in January 1994, settled: according to Kirk McKusick, three files of about 18 thousand were removed, and around seventy received copyright notices from USL, with free redistribution. In October 1993, Novell transferred the UNIX trademark to X\/Open, separating the name from the code of any company.<\/p>\n

1991: Linux grows while BSD is in court<\/h2>\n

On 25 August 1991, Linus Torvalds, a student in Helsinki, announced in the group comp.os.minix<\/code> that he was writing a free operating system \u201cjust for fun\u201d. His working environment was MINIX<\/strong>, a Unix-like teaching system created by Andrew Tanenbaum in 1987 to teach operating systems. Linux did not copy MINIX's code: it is a new kernel, written for the 386, that used the tools from the GNU<\/strong>, project, by Richard Stallman, to form a complete system.<\/p>\n

The timing helped. Free BSD was stuck in the process, and the GNU kernel, the Hurd, wouldn't get ready. Linus himself said, years later, that if 386BSD or the Hurd had been available at the time, he might never have written Linux. Linux isn't a certified Unix, but it follows POSIX and inherited the ideas, the commands, and the culture. The complete story is in The history of Linus Torvalds<\/a> and in Linux at 35<\/a>.<\/p>\n

1995\u20132010: SCO, Solaris, and the Unix that became Mac<\/h2>\n

In 1995, Novell sold the UnixWare business to Santa Cruz Operation; in 1996, X\/Open and OSF merged into the The Open Group<\/strong>, which manages the trademark to this day. In 2002, Caldera, which had bought SCO's Unix division, released the source code for editions 1 through 7 and for 32V under a BSD-style license. That's why anyone can download and run these old Unixes legally.<\/p>\n

Sun took Solaris to its peak. The Solaris 10<\/strong>, from 2005, brought DTrace and Zones, and ZFS arrived in an update in 2006. In 2010, Oracle bought Sun and shut down OpenSolaris; the community kept the open source code going in illumos<\/strong>, based on OpenIndiana, OmniOS, and SmartOS.<\/p>\n

Unix also reached the desktop through an unexpected path. NeXTSTEP, by Steve Jobs, with a kernel based on Mach and parts of BSD, became Mac OS X in 2001. The kernel is called XNU<\/strong>, an acronym for \u201cX is Not Unix,\u201d but the Mac OS X 10.5 Leopard<\/strong>, in 2007, was certified as UNIX 03, and macOS has kept the seal ever since. The open part of the system is called Darwin. Apple's turnaround lies in Steve Jobs history<\/a>.<\/p>\n

In the midst of this, the SCO Group sued IBM, Novell and Linux users claiming rights over Unix code present in Linux. In March 2010, a jury decided that Unix copyrights belonged to Novell, and SCO's thesis collapsed. UnixWare and OpenServer today belong to Xinuos.<\/p>\n

2025: the lost tape of the fourth edition<\/h2>\n

The Utah tape closes a gap in history. It was among materials of professor Jay Lepreau, who died in 2008, and was found by Aleksander Maricq, a researcher at the Flux Research Group, while cleaning a storage room in the Merrill Engineering Building in 28 July 2025. Professor Robert Ricci announced the discovery in November, and the tape was taken to the Computer History Museum, in California.<\/p>\n

In December, Al Kossow, the museum's software curator, performed the reading: instead of reading the bytes directly, the equipment recorded the raw magnetic signal, and the content was reconstructed from it. Only two blocks failed, and there was enough information to recover them. The image is on Internet Archive<\/a>, and the fourth edition, the first with a kernel in C, runs again on the SimH emulator.<\/p>\n

Who is still Unix in 2026<\/h2>\n

\u201cUNIX\u201d is a trademark of The Open Group, and only those who pass the Single UNIX Specification certification may use it. In the official registry<\/a>, as of September 2026, the systems certified under the most recent standards are:<\/p>\n