Unix and Linux History: From 1969 Reach Now
See how Bell Labs start Unix in 1969, AT&T licensing spread am, BSD rise, the 1992 lawsuit land, and Linux finally pass BSD for server rooms.
Unix and Linux history, short version
Unix and Linux history na one long argument about who get right to own the source code. Unix start for Bell Labs for 1969. Linux start for Helsinki for 1991, and e no share any of the original code. Wetin move between both na design and one set of published interfaces: files, processes, pipes, and shell wey join small programs together. Unix spread through universities for the 1970s because one 1956 antitrust settlement stop AT&T (American Telephone and Telegraph) from selling software, so Bell Labs license the source cheaply instead. Unix later dey everywhere, even though nobody outside AT&T own am, and the licensing fights wey follow decide which free Unix-like system finally enter your server rack.
1969: one spare PDP-7, and the ideas wey remain
Bell Labs comot for the Multics project for 1969. Multics (multiplexed information and computing service) na big time-sharing system wey dem build with MIT and General Electric, and the Labs decide say e too big to finish. Ken Thompson keep the ideas wey e like and drop the rest, then write small time-sharing system for one cast-off PDP-7 minicomputer. Dennis Ritchie join am. The name Unix na joke against Multics.
For 1970, the Labs order PDP-11 because dem promise system for editing and formatting text. Text processing tool dey easier to fund pass operating system, and typists for the patent department become the first users. Unix dey run for PDP-11 by February 1971. That machine get 24 KB core memory, wey dem divide between the system and user programs. This funding accident na why the first Unix manual, dated November 1971, be set of formatted documents. Na why you still dey type man 5 crontab. The section numbers for that manual na the same section numbers for your own.
Two changes make the design remain permanent. Pipes enter for Version 3 in 1973, because Doug McIlroy don argue for years say programs suppose fit join end to end. Thompson add the | operator, so output from one program become input for the next program. Then Version 4, later for 1973, dem rewrite am for C. Operating system wey dem write for portable language fit move go hardware wey dem no write am for. Na why Unix outlive every machine wey e start on.
Why Unix spread: AT&T no fit sell am
A 1956 consent decree settle antitrust case against AT&T. AT&T keep im telephone monopoly and accept one limit in return: e go stay away from businesses wey no be telecommunications. Software na one of those businesses. So when universities ask for Unix, Bell Labs no fit sell am as product. E license the source for small fee, without support and without warranty.
The effect big, and e no be wetin AT&T plan. Sixth Edition Unix, released for 1975, reach hundreds of computer science departments with full source. John Lions for University of New South Wales print that kernel source with line-by-line commentary and teach from am. One generation learn how operating system dey work by reading real one.
Then the license change. The Seventh Edition license for 1979 forbid people from using the source for classes, so the Lions commentary circulate as photocopies of photocopies. Na so the whole story be. Unix dey everywhere and e no free at the same time, so every improvement wey anybody make become improvement to another person proprietary code.
Berkeley: the parts of Unix wey you actually dey type
Ken Thompson spend the 1975 to 1976 academic year for University of California, Berkeley, and leave one Unix group wey dey very active behind. Berkeley Computer Systems Research Group (CSRG) release tapes wey carry local improvements, and those tapes later become BSD, the Berkeley Software Distribution. Bill Joy write plenty of the early work: 1BSD for 1978, then 2BSD for 1979 wey carry the vi editor and the C shell.
Na from the C shell !! and !$ come. Bash inherit that syntax, na why bash history expansion still dey surprise people wey type exclamation mark inside double quotes even forty-something years later.
DARPA (the US Defense Advanced Research Projects Agency) later pay Berkeley to put the new internet protocols inside Unix. 4.2BSD, wey release for August 1983, ship TCP/IP (transmission control protocol over internet protocol) and the socket application programming interface. Every TCP service for your server, from sshd reach your web server, still call socket(), bind(), listen() and accept() for that order, because na Berkeley pick those names for 1983.
One detail decide wetin go happen for the next decade. BSD tape no be complete system. Na set of additions to AT&T Unix, and you need valid AT&T source license to run am legally. Berkeley replace the AT&T parts with its own code over several years, file by file. Whether that replacement truly complete na the question wey later reach court.
1984: breakup, then Unix wars
Bell System dem fashi am for 1 January 1984, and consent decree wey bin stop AT&T from entering software business go with am. AT&T fit now sell Unix as product, and e do am. Commercial source licenses become expensive, so Unix stop to be the cheap thing wey university dey give students.
The vendors don already fork am. Every workstation company dey ship im own Unix for im own hardware, so by late 1980s, program wey dem write for one system need porting before e fit work for another. The industry split into two standards camps for 1988: Open Software Foundation for one side and Unix International for the other. Dem spend years dey ship systems wey no compatible, while dem dey argue about which Unix be the real one.
POSIX (portable operating system interface) come out from that wahala. IEEE 1003.1 publish for 1988, and e write down the system calls and C library interfaces wey Unix-like system must provide. The shell and standard utilities follow for separate standard, IEEE 1003.2, for 1992. This matter pass how e sound, because standard na specification and no license cover am. Three years after the first POSIX standard, one student write kernel against those documents.
Minix, and the gap wey e leave
Andrew Tanenbaum release Minix for 1987 as the teaching system for e operating systems textbook. Minix na small Unix-like system; the source dey come with the book, and e run for the cheap PCs wey students actually get. Person fit legally read am for classroom, unlike Seventh Edition Unix, wey no longer allow that.
Minix remain small on purpose, because person suppose fit explain am for book, and Tanenbaum reject patches wey for turn am into production system. The license no free too: you buy the book, and na no be your decision to redistribute your modified Minix. So by 1991, student fit read working Unix-like kernel, but e no fit build anything wey go last on top am.
GNU get everything except kernel
Richard Stallman announce GNU project for September 1983, with aim to build complete free Unix-like system. By 1991 GNU don produce most of the parts wey dey around kernel: GCC compiler, GNU C library, binary utilities, and bash, the shell wey Brian Fox write for 1989 and wey your server still dey start you inside. GNU kernel, Hurd, na the part wey keep slipping.
Dem publish GNU General Public License version 2 for June 1991. The rule short. You fit use the code and change am, but if you distribute the result, you must make the source available under the same terms, either alongside the program or through written offer to supply am. That rule go become the hinge of this story two sections from now.
August 1991: post wey go comp.os.minix
For 25 August 1991, one student for Helsinki post message for Usenet group comp.os.minix:
Hello everybody out there using minix -
I'm doing a (free) operating system (just a hobby, won't be big and
professional like gnu) for 386(486) AT clones.Version 0.01 come out for September 1991. E no share code with Unix or Minix. Na new kernel for 386, wey dem develop for Minix machine, write against POSIX interfaces, and pair with GNU tools wey don already dey available. For January 1992, Tanenbaum tell Torvalds say monolithic kernel na obsolete design. E get valid point about design, but Torvalds dey answer different question: wetin fit run well for the one cheap computer wey student get.
For February 1992, dem relicense version 0.12 under GPL. Torvalds don later call am im best decision. The original Linux license no allow money to change hand, and this one for block the CD vendors and support companies wey come after. GPL allow commerce, but e force every distributed change to go back into the same shared tree.
The lawsuit: April 1992 to February 1994
Berkeley release Networking Release 2 (Net/2) for June 1991: e nearly complete BSD system, with the files wey come from AT&T don remove. Six kernel files no dey. Bill and Lynne Jolitz write replacements and release 386BSD 0.0 for March 1992, then 386BSD 0.1 for 14 July 1992. Free, complete, mature BSD for the 386 don dey available, for the same time Linux still be hobby kernel. Berkeley Software Design, Inc. (BSDi) sell supported commercial build, BSD/386, and advertise am with telephone number 1-800-ITS-UNIX.
For April 1992, Unix System Laboratories (USL), the AT&T subsidiary wey own Unix at that time, sue BSDi for federal court for New Jersey over trade secrets and the Unix trademark. Dem later amend the lawsuit to add the Regents of the University of California. The University countersue for California for 1993, dey argue say AT&T ship Berkeley code inside System V without the credit wey its own license require.
The legal core scatter for 1993. Judge Dickinson Debevoise refuse USL preliminary injunction, because e believe say USL copyright claim over the older 32V code probably no valid. The reason be say AT&T distribute that code for years without copyright notices. Novell buy USL from AT&T for the middle of 1993, and Novell management want make the fight end. The case settle for February 1994. Out of about 18,000 files for the Berkeley distribution, dem remove three, and about seventy receive USL copyright notices.
Berkeley ship 4.4BSD-Lite for June 1994, and e legally clean. FreeBSD and NetBSD, wey both start for 1993 on the disputed Net/2 code, then need rebuild their systems on the new base. The rebuild take most of the remaining 1994. Linux 1.0 release for March 1994, for the middle of that rebuild.
Why Linux take over the server market and BSD no?
The lawsuit na the popular answer, and e dey part of the answer. Between April 1992 and February 1994, anybody wey choose free Unix for product had to weigh an active claim from AT&T lawyers against kernel from Finnish student wey nobody fit sue over. Na that exact period web show up and first Linux distributions appear: Slackware for July 1993 and Debian for August 1993, while Red Hat and SUSE follow soon after. Red Hat turn that early advantage into the commercial standard, and the journey from Red Hat Linux to CentOS and then Rocky and AlmaLinux explain why distribution wey first release for 1995 still decide wetin dey run for plenty business servers.
Four other things matter as much as the court case.
- Hardware. Linux target the commodity 386 PC from the first line of code, and na that hardware become cheap. BSD main focus na VAX and workstation, while the 386 port na outside effort from two people.
- The license. GPL require company wey dey ship modified kernel to publish its changes, so vendor work flow back into one tree. BSD license allow company keep its changes private, and companies actually keep dem.
- The development model. Torvalds merge patches from strangers quickly and release constantly. 386BSD release slowly enough that its own users fork am twice for 1993, into NetBSD and FreeBSD, and again into OpenBSD for 1995.
- Momentum. Developers go where other developers already dey, and drivers dey get written for the system wey get the most users.
Make we dey honest about the technical question. BSD for 1994 na the more complete system, with coherent base, documented history, and networking code wey Linux spend years trying to catch up with. Nobody choose Linux for 1994 because e better. People choose am because e dey available, e no get legal restriction, e dey run on hardware wey dem already own, and e dey improve every week.
BSD wetin retain, and where e dey run now
BSD continue to dey. Wetin e lose na the default position. The clearest evidence dey your own machine: OpenBSD project write OpenSSH for 1999, and na the SSH server for almost every Linux system wey ship today. Na why SSH key habits wey worth learning dey the same for both families.
Netflix dey serve video from FreeBSD appliances. Junos OS, wey be the operating system for Juniper routers, get FreeBSD underneath, although Juniper newer Junos OS Evolved dey run on Linux. The PlayStation system software come from FreeBSD. Apple macOS and iOS carry BSD code inside the kernel and across the userland. The permissive license wey make BSD lose the shared server tree put BSD code inside plenty hardware wey no ever mention am.
If you dey choose today, the question practical pass historical one. Comparing Linux and FreeBSD as a server dey mainly depend on ZFS, jails, the ports tree, and how much third party software assume Linux underneath. FreeBSD 15 as a server na current system wey dem still dey maintain, no be museum exhibit. And Linux against Windows Server na separate question, and the answer depend on your application stack.
The 1969 design wey you dey use for VPS today
Each one of these old pass most of the people wey dey use am.
- The pipe.
who | wc -ldey count logged-in users because Version 3 Unix for 1973 allow output from one program become input for another program. - The manual sections.
man 1 lsandman 5 crontabdey use numbering scheme from the November 1971 manual. - The filesystem split.
/usrdey exist because root disk for that PDP-11 full up for 1971, so the developers move files go the second disk pack. Distributions don later undo the split: runls -ld /binfor Ubuntu 24.04 or Debian 13 and you go get symbolic link tousr/bin. - The socket calls. Berkeley write dem for 4.2BSD for 1983, and every network daemon wey dey accept TCP connections still dey use dem.
- POSIX. A
#!/bin/shscript wey dem write against the standard go run unchanged for Linux, FreeBSD, macOS, and Solaris, because na the standard all of dem implement.
One inheritance no survive the journey: the shell-script boot sequence wey Linux take from System V don disappear from nearly every distribution now, and the argument wey replace am with systemd na the most recent fight about wetin Unix-like system suppose be.
When you rent virtual private server and log in, na interface wey dem design to serve machine wey get 24 KB of memory and patent department you dey type for. E survive because dem publish the interface, people argue about am, dem standardise am, and later people wey no ever get permission to see the original code reimplement am from scratch.
FAQ
Linux get any original Unix code inside?
No. Linus Torvalds write the kernel from scratch from 1991, and e target the POSIX interfaces instead of any AT&T source. The Unix inheritance na design and published interface: files, processes, pipes, and the system call names. Dem write the GNU tools around the kernel from scratch too, starting from 1983. Na this independence make the AT&T litigation over BSD code no touch Linux at all.
When AT&T lawsuit against BSD happen, and e kill BSD?
USL, the AT&T subsidiary wey own Unix, sue BSDi for April 1992 and later add the University of California as defendant. For 1993, the judge refuse preliminary injunction, and rule say the copyright claim over the older 32V code likely no valid. Novell buy USL for middle of 1993 and settle the case for February 1994. Dem remove three files from about 18,000, and give roughly seventy new copyright notices. The case no kill BSD. E freeze BSD adoption during the two years wey the world dey choose free Unix.
Why Linux take over servers instead of BSD?
Legal certainty na one reason: through 1992 and 1993, Linux carry no lawsuit while BSD carry one. Other reasons na 386 as target from day one, GPL wey pull vendor changes back into one tree, and merge process wey fast enough to keep contributors interested while 386BSD fork three ways. Technical merit no be the deciding factor, because BSD na the more finished system for 1994.
FreeBSD still worth running today?
Yes, and na concrete reasons: ZFS integrate with the base system, jails as mature isolation model, base system wey dem develop as one coherent unit, and documentation wey remain accurate. The cost na compatibility work, because most third party server software, container tooling, and vendor support assume Linux. Choose FreeBSD when e storage and networking work justify that cost, instead of choosing am because of loyalty to the older lineage.