1.3. UNIX. A bit of history

As a predecessor to our GNU/Linux systems [Sta02], let's recall a bit about the history of UNIX [Sal94] [Lev]. Originally, Linux was conceived as a Minix clone (an academic implementation of UNIX for PC) and used some ideas developed in proprietary UNIX; but, in turn, it was developed in open source, and with a focus on domestic PCs. In this section on UNIX and in the following one on GNU/Linux, we will see how this evolution has brought us to current day GNU/Linux systems that are capable of competing with any proprietary UNIX and that are available for a large number of hardware architectures, from the simple PC to supercomputers.

Linux can be used on a broad range of machines. In the TOP500 list, we can find several supercomputers with GNU/Linux (see list on webpage top500.org): for example, the MareNostrum, in the Barcelona Supercomputing Center, a cluster, designed by IBM, with 10240 CPUs PowerPC with GNU/Linux operating system (adapted to the requirements of these machines). From the list we can see that overall supercomputers with GNU/Linux make up 75% of the list.

Example 1-7. Note

We can see the TOP500 list of the fastest supercomputers at:


UNIX started back in 1969 (we now have almost 40 years of history) in the Bell Telephone Labs (BTL) of AT&T. These had just withdrawn from a project called MULTICS, which was designed to create an operating system so that a large computer could support thousands of users simultaneously. BTL, General Electric, and MIT were involved in the project. But it failed, in part, because it was too ambitious for the time.

While this project was underway, two BTL engineers who were involved in MULTICS: Ken Thompson and Dennis Ritchie, found a DEC PDP7 computer that nobody was using, which only had an assembler and a loading program. Thompson and Ritchie developed as tests (and often in their free time) parts of UNIX, an assembler (of machine code) and the rudimentary kernel of the operating system.

That same year, in 1969, Thompson had the idea of writing a file system for the created kernel, in such a way that files could be stored in an ordered form in a system of hierarchical directories. Following various theoretical debates (which took place over about two months) the system was implemented in just a couple of days. As progress was made on the system's design, and a few more BTL engineers joined in, the original machine became too small, and they thought about asking for a new one (in those days they cost about 100,000 US dollars, which was a considerable investment). They had to make up an excuse (since the UNIX system was a free time development) so they said they wanted to create a new text processor (an application that generated money at that time), so they were given approval to purchase a PDP11.


UNIX dates back to 1969, with over 30 years of technologies developed and used on all types of systems.

When the machine arrived, they were only given the CPU and the memory, but not the disk or the operating system. Thompson, unable to wait, designed a RAM disk in memory and used half of the memory as a disk and the other half for the operating system that he was designing. Once the disk arrived, they continued working on both UNIX and the promised text processor (the excuse). The text processor was a success (it was Troff, an editor language subsequently used for creating the UNIX man pages), and BTL started using the rudimentary UNIX with the new text processor, with BTL thus becoming the first user of UNIX.

At that time, the UNIX philosophy started to emerge [Ray02a]:

Another important characteristic was that UNIX was one of the first systems conceived to be independent of the hardware architecture, and this has allowed it to be carried over to a large number of different hardware architectures.

In November 1971, as there were external users, the need to document what was being done resulted in the UNIX Programmer's Manual signed by Thompson and Richie. In the second edition (June 1972), known as V2 (the edition of the manuals was made to correspond with the UNIX version number), it was said that the number of UNIX installations had already reached 10. And the number continued to grow to about 50 in V5.

Then, at the end of 1973, it was decided to present the results at a conference on operating systems. And consequently, various IT centres and universities asked for copies of UNIX. AT&T did not offer support or maintenance to UNIX, which meant that users had to unite and share their knowledge by forming communities of UNIX users. AT&T decided to cede UNIX to universities, but did not offer them support or correct errors for them. Users started sharing their ideas, information on programs, bugs etc. They created an association called USENIX, meaning users of UNIX. Their first meeting in May 1974 was attended by a dozen people.

Example 1-8. Note

See: http://www.usenix.org

One of the universities to have obtained a UNIX license was the University of California at Berkeley, where Ken Thompson had studied. In 1975, Thompson returned to Berkeley as a teacher bringing with him the latest version of UNIX. Two recently-graduated students, Chuck Haley and Bill Joy (nowadays one of the vice-presidents of SUN Microsystems), joined him and started to work together on a UNIX implementation.

One of the first things that they were disappointed with were the editors; Joy perfected an editor called EX, until transforming it into VI, a full screen visual editor. And the two developed a Pascal language compiler, which they added to UNIX. There was a certain amount of demand for this UNIX implementation, and Joy started to produce it as the BSD, Berkeley Software Distribution (or UNIX BSD).

BSD (in 1978) had a particular license regarding its price: it said that it corresponded to the cost of the media and the distribution it had at that time. Thus, new users ended up making some changes or incorporating features, selling their remade copies and after a certain amount of time, these changes were incorporated into the following version of BSD.

Joy also made a few more contributions to his work on the vi editor, such as handling text terminals in such a way that the editor was independent of the terminal where it was being used; he created the TERMCAP system as a generic terminals interface with controllers for each specific terminal, so that programs could be executed irrespective of the terminals using the interface.

The following step was to adapt it to different architectures. Until 1977, it could only be run on PDP machines; that year adaptations were made for machines of the time such as Interdata and IBM. UNIX Version 7 (V7 in June 1979) was the first portable one. This version offered many advances, as it included: awk, lint, make, uucp; the manual already had 400 pages (plus two appendices of 400 pages each). It also included the C compiler designed at BTL by Kernighan and Ritchie, which had been created to rewrite most of UNIX, initially in the assembler and then into C with the parts of the assembler that only depended on the architecture. Also included were an improved shell (Bourne shell) and commands such as: find, cpio and expr.

The UNIX industry also started to grow, and versions of UNIX (implementations) started to appear from companies such as: Xenix, a collaboration between Microsoft – which in its early days it also worked with UNIX versions – and SCO for Intel 8086 machines (the first IBM PC); new versions of BSD from Berkeley...

However, a new problem appeared when AT&T realised that UNIX was a valuable commercial product, the V7 license prohibited its study in academic institutions in order to protect its commercial secret. Until that time many universities used the UNIX source code in order to teach operating systems, and they stopped using it to teach only theory.

However, everyone found their own way of solving the problem. In Amsterdam, Andrew Tanenbaum (prestigious author of theory books on operating systems) decided to write a new UNIX-compatible operating system without using a single line of AT&T code; he called this new operating system Minix. This is what would subsequently be used in 1991 by a Finnish student to create his own version of UNIX, which he called Linux.

Bill Joy, who was still at Berkeley developing BSD (already in version 4.1), decided to leave to a new company called SUN Microsystems, where he finished working on BSD 4.2, which would later be modified to create SUN's UNIX, SunOS (around 1983). Every company started developing its own versions: IBM developed AIX, DEC - Ultrix, HP - HPUX, Microsoft/SCO - Xenix etc. As of 1980, UNIX began as a commercial venture, AT&T released a final version called UNIX System V (SV), on which as well as on the BSD 4.x, current UNIX are based, whether on the BSD or the System V branch. SV was revised several times and, for example, SV Release 4 was one of the most important ones. The result of these latest versions was that more or less all existing UNIX systems were adapted to each other; in practice they are versions of AT&T's System V R4 or Berkeley's BSD, adapted by each manufacturer. Some manufacturers specify whether their UNIX is a BSD or SV type, but in reality they all have a bit of each, since later several UNIX standards were drawn up in order to try and harmonise them; among these, we find IEEE POSIX, UNIX97, FHS etc.

Over time, the UNIX system split into several branches, of which the two main ones were AT&T's UNIX or System V, and the University of California's BSD. Most current UNIX systems are based on one or the other, or are a mixture of the two.

However, at that time, AT&T (SVR4) was undergoing legal proceedings as a telephone monopoly (it was the leading, if not the only, telephone company in the US), which forced it to split into several smaller companies, causing the rights to UNIX to start dancing between owners: in 1990 it was shared 50/50 by the Open Software Foundation (OSF) and UNIX International (UI), later, UNIX Systems Laboratories (USL), which denounced the University of Berkeley for its BSD copies, but lost, since the original license did not impose any ownership rights over the UNIX code. Later, the rights to UNIX were sold to Novell, which ceded a share to SCO, and as of today it is not very clear who owns them: they are claimed through different fronts by Novell, the OSF and SCO. A recent example of this problem is the case of SCO, which initiated a lawsuit against IBM because according to SCO, it had ceded parts of the UNIX source code to versions of the Linux kernel, which allegedly include some original UNIX code. The result as of today is that the matter remains in the courts, with SCO turned into a pariah of the IT industry threatening Linux, IBM, and other proprietary UNIX users, with the assertion that they own the original rights to UNIX and that everyone else should pay for them. We will have to see how this case evolves, and the issue of UNIX rights along with it.

The current scenario with UNIX has changed a lot since Linux appeared in 1991, since as of 1995-99 it became a serious alternative to proprietary UNIX systems, due to the large number of hardware platforms that it supports and the extensive support for its progress of the international community and companies. Different proprietary versions of UNIX continue to survive in the market, because of their adaptation to industrial environments or for being the best operating system in the market, or because there are needs that can only be covered with UNIX and the corresponding hardware. Also, some proprietary UNIX are even better than GNU/Linux in terms of reliability and performance although the gap is shortening all the time, since companies with their own proprietary UNIX systems are showing more and more interest in GNU/Linux and offering some of their own developments for inclusion in Linux. We can expect a more or less slow extinction of proprietary UNIX versions towards Linux-based distributions from manufacturers adapted to their equipment.

Overview of these companies:

Example 1-9. Note

Many companies with proprietary UNIX participate in GNU/Linux and offer some of their developments to the community.

Example 1-10. Note

Open letter from Novell to the GNU/Linux community

http://www.novell.com/linux/microsoft/community_open_ letter.html

Another interesting historical anecdote is that together with a company called UniSys, they launched a marketing campaign on how to convert UNIX systems to Windows systems; and although its purpose may be more or less commendable, a curious fact is that the original web server of the business was on a FreeBSD machine with Apache. Occasionally, it also pays "independent" companies (some would say they are not very independent) to conduct comparative performance analyses between UNIX/Linux and Windows.


As a general summary, some comments that tend to appear in UNIX bibliography point to the fact that UNIX is technically a simple and coherent system designed with good ideas that were put into practice, but we should not forget that some of these ideas were obtained thanks to the enthusiastic support offered by a large community of users and developers who collaborated by sharing technology and governing its evolution.

And since history tends to repeat itself, currently that evolution and enthusiasm continues with GNU/Linux systems.