Two weeks ago we mapped where Unix lives: macOS on hundreds of millions of desks, FreeBSD under a substantial share of evening internet, BSD in the consoles, BSD in the routers, AIX and Solaris and HP-UX and z/OS in the enterprise back office. Last week we drew the line: certified Unix on the trademark register, descended-uncertified BSDs by inheritance, Unix-like Linux by re-implementation, with QNX importing the BSD network stack and sitting near the family without claiming the name.
This week, the source. Where did this family come from, how did it spread to half the world's running infrastructure, and what makes its fifty-seven-year arc unusual in the history of software. The story is shorter than the influence suggests, and the people in it are fewer than one would expect.
Bell Labs, 1969
In the late 1960s, AT&T's Bell Telephone Laboratories was a participant in Multics, an ambitious time-sharing operating system being built jointly with MIT and General Electric. The project, by 1969, was overrun: too many goals, too much complexity, too few delivered features. Bell Labs withdrew that year, having reached the unsurprising conclusion that an operating system nobody can finish is of limited practical use.
One of the engineers who came back to Bell Labs from the Multics work was Ken Thompson. He found a little-used DEC PDP-7 minicomputer going spare in a corner of the lab and began writing, partly as an exercise, partly out of stubbornness, a much smaller system. The file system came first. Then a shell. Then a few utilities. The system did not have a name yet.
A colleague, Brian Kernighan, by his own later account suggested calling it Unics, a pun on the over-ambitious Multics it was meant to do without; the spelling soon settled as Unix. The name stuck.
Two more engineers joined the work: Dennis Ritchie and Douglas McIlroy. McIlroy, in particular, contributed the design idea that became the system's signature: small programs that did one thing well, composed through pipes, communicating in plain text. The Unix philosophy was not a manifesto. It was a working style that emerged from the constraints of the PDP-7.
In 1973, the system was rewritten in C, a language Dennis Ritchie had developed for the purpose. This was the moment that made everything afterwards possible. An operating system written in a high-level language could, in principle, be moved to any computer with a C compiler. Unix became portable. Almost nothing else of its generation was, and the advantage this bought over the next two decades was not a small one.
The Spread, 1973–1980
Bell Labs, by US antitrust agreement, was not allowed to enter the computer business. AT&T could not sell Unix as a product. What it could do, and did, was license the source code to universities for a nominal fee.
The universities ran with it. Version 6 (V6), released in 1975, was the first version to spread widely outside Bell Labs. Lions' Commentary on UNIX 6th Edition, a line-by-line reading of the kernel source written by John Lions at the University of New South Wales in 1976, became one of the most widely photocopied computer-science documents of the era. AT&T legal departments later restricted its distribution; samizdat versions circulated for two decades.
In 1977, the Computer Systems Research Group at the University of California, Berkeley, began distributing its own modifications to Unix under the name Berkeley Software Distribution (BSD). What had begun as patches and additions became, over the next fifteen years, a fully alternative line: 1BSD, 2BSD, 3BSD, 4BSD, with each release adding components that AT&T eventually adopted back. TCP/IP networking, in 4.2BSD (1983), is the obvious example: every modern operating system on the internet uses what Berkeley wrote.
By the early 1980s, there were two Unix lines in circulation: AT&T's commercial System V (released 1983), descended from the Bell Labs research line, and the BSD line from Berkeley, descended from the same root but diverged in code and philosophy. Both were Unix. Both had the same ancestor. Their separation was real.
The Commercial Era, 1980–1990
Through the 1980s, Unix became the lingua franca of the workstation market. Sun Microsystems built SunOS on a BSD base. DEC, IBM, HP, Silicon Graphics, NeXT and a dozen others each shipped their own Unix flavour for their own hardware: Ultrix, AIX, HP-UX, IRIX, NeXTSTEP. Each was Unix; each was incompatible in irritating small ways. The market called this period the Unix wars.
The fragmentation produced its own remedy. In 1988, IEEE published POSIX (1003.1-1988), a standard for the API rather than the implementation. The same set of system calls would be available across the divergent flavours, even if the underlying kernels differed. POSIX did not end the fragmentation, but it made portable software possible across it.
In 1989, the rights to the original Unix source code passed from AT&T to its Unix System Laboratories (USL) subsidiary, and from USL through a complicated sequence (Novell 1993, X/Open 1994, The Open Group 1996) to where the trademark sits today.
The Lawsuit, 1992–1994
In 1991, Berkeley released Net/2, an attempt to ship a Unix distribution that contained no AT&T-licensed code, only Berkeley-written code. Berkeley Software Design Inc. (BSDi), a commercial spinoff, began selling a product based on Net/2 in 1992.
USL sued. The argument was that some AT&T-derived code remained in Berkeley's release. The case dragged through 1992 and 1993, with the discovery process forcing Berkeley to publish, line by line, what was original and what was derived. The settlement, in 1994, required Berkeley to remove three files and add copyright notices to roughly seventy more, out of some eighteen thousand, and let everything else stand. It was not the decisive vindication AT&T's lawyers had been retained to deliver.
The result was 4.4BSD-Lite, the clean Berkeley release. From it descend, directly, FreeBSD (1993), NetBSD (1993), OpenBSD (1996, forked from NetBSD), and DragonFly BSD (2003). The BSD lineage that runs on Netflix, Sony PlayStation, Juniper, and inside macOS's userland is the lineage that survived 1994.
Free Unix, 1991
1991 is the year Unix itself became free. Berkeley shipped Net/2 in mid-1991, the first release substantially clear of AT&T code; from it came 386BSD in 1992 and, once the litigation settled, FreeBSD and NetBSD in 1993, OpenBSD and DragonFly later. This is Unix in the full sense the second piece set out: descended in its own source from Bell Labs and Berkeley, carrying the lineage rather than merely resembling it. The free Unix that runs today is this line.
The same year, and as a separate story, something else began. A student in Helsinki, Linus Torvalds, had been working on MINIX, the small teaching Unix that Andrew Tanenbaum had written in 1987. He liked Unix a great deal and found MINIX too confining: it was kept deliberately simple for the classroom and licensed too tightly to extend freely. So he started a kernel of his own, from scratch, with no Bell Labs or Berkeley code in it at all. Paired with the GNU userland that Richard Stallman's project had been building since 1983, it became GNU/Linux, and reached its first stable release, 1.0, in 1994, a year after FreeBSD's own.
These are not two versions of one thing, and it is worth not blurring them. The BSDs are Unix by descent. Linux is Unix-like by resemblance: a unixoid, a fresh re-implementation that took the design as inspiration and inherited none of the code. That it later became the most widely deployed Unix-like kernel on Earth is a fact about reach, and reach is not lineage. In the history of Unix proper, Linux is the parallel line that runs alongside, not a second line within it.
What the Genealogy Explains
The fifty-seven-year arc of Unix is the result of four moments between 1969 and 1994, each of them small in scale, each of them with consequences out of proportion to their size: a discarded PDP-7 in a corner at Bell Labs; a licence agreement that pushed the source into universities; a research group at Berkeley that built an alternative line; a lawsuit that ended with that line set free.
None of the people involved were trying to take over the world. None of the institutions involved were positioned to monetise what was happening. The economics of Unix's spread are unusual in the history of software: the value was created by gift-economy circulation, by source-code reading in university labs, by collaborative debugging across the early internet, and only later captured by the commercial layer that made workstations and enterprise software.
Unix is what happens when an operating system is small enough to read, portable enough to move, and licensed loosely enough that the people who care can extend it. It is not a brand and not a single project. It is the family that came out of those moments: the BSDs as one branch, the certified commercial vendors as another. Linux stands beside the family rather than within it, a parallel re-implementation and a unixoid rather than a Unix; QNX and a handful of others import pieces under their own names.
This is a family tree, not the only one that could be drawn. A separate engineering tradition, the microkernel line of MINIX, Mach and L4, grew up beside it with a genealogy of its own, and deserves its own telling rather than a footnote here.
The free BSD line and the free Linux kernel are exactly the same age, both 1991. The difference that actually matters is not in the calendar but in the code: the BSDs are Unix by descent, Linux is Unix-like by resemblance.
The Point
Where Unix runs today, the first piece showed: half the world's infrastructure, almost never advertising itself. What Unix is, the second piece drew: a trademark, a specification, a register, plus the family of relatives that sit just outside the line. Where Unix came from, this piece has tried to tell: Bell Labs 1969, Berkeley 1977, System V 1983, the clean Berkeley release of 1994. Fifty-seven years. Hundreds of millions of running systems.
It is unlikely that any operating-system family today, given the venture capital and the lock-in economics of the modern industry, could repeat the trajectory. Unix is the artefact of a particular moment in computing, when computers were rare enough to be read carefully, source code was treated as scholarly material, and the contributors knew each other by name.
The family that emerged has outlived the moment. It is still running. It is, in the precise sense of the word, quiet, everywhere, reliable, certified: all four. This bow ends here. The family is what we got.