The Past and Future of Linux Standards

“The nice thing about standards is that there are so many of them to choose from.” --Professor Andrew S. Tanebaum (author of MINIX).

Despite their well-earned reputation as a source of confusion, standards are one of the enabling factors behind the success of Linux. If it were not for the adoption of the right standards by Linus Torvalds and other developers, Linux would likely be a small footnote in the history of operating systems.

The Early Days

Some people believe the interest in Linux standards is very recent, precipitated by the upswing in commercial interest. However, before Linux was even named, conformance to an open standard was an important goal. Here is one of the first postings by Linus Torvalds about a project that would soon be named Linux:

From: torvalds@klaava.Helsinki.FI
        (Linus Benedict Torvalds)
Newsgroups: comp.os.minix
Subject: Gcc-1.40 and a POSIX-question
Date: 3 Jul 91 10:00:50 GMT
Due to a project I'm working on (in minix),
I'm interested in the POSIX standard definition.
Could somebody please point me to a (preferably)
machine-readable format of the latest POSIX rules?
FTP sites would be nice.

A month later, Linus posted:

As to POSIX, I'd be delighted to have it,
but POSIX wants money for their papers, so that's
not currently an option.
Despite the high cost of a copy of the POSIX standard in 1991, it became one of the primary standards for Linux. POSIX, the Portable Operating System Interface, is a standard application programming interface (API) used by Linux and many other operating systems (typically UNIX and UNIX-like systems). There are several major benefits to using the interface defined by POSIX. It makes it easier to write source code that can be compiled on different POSIX systems. It also gives Linux application developers and Linux kernel developers a well-defined API to share. That means application developers don't need to track most kernel changes as long as the kernel continues to behave as POSIX says it should.

In addition, using POSIX as the API for Linux enabled Linus and other early Linux developers to use existing free programs written by the GNU Project, the BSD operating system and many other free programs which adhere to the POSIX specification.

It is important to note that POSIX does not provide for precompiled binary applications to be run on any POSIX operating system. Since it provides source code compatibility, but not binary compatibility, POSIX is often thought of as a source standard. Early development of Linux was made under the Minix operating system. In fact, Linus originally wanted to make Linux binary-compatible with Minix. That idea was dropped when the differences between Linux and Minix became too great, but some traces of the Minix heritage of Linux still linger here and there.

When asked about the importance of POSIX in early Linux development, Linus Torvalds said, “Linux started out very aware of POSIX, but even more so of the unofficial de facto standards.” He elaborated that, at the time, these de facto standards were approximately SunOS (the precursor to Solaris) behavior, somewhere between BSD and System V. “Basically, I wanted to not have to spend too much time porting user mode programs (it wasn't something I was all that interested in), and POSIX helped in that,” Torvalds said.

Linux Today

POSIX.1 (the POSIX kernel interface) is still considered by Linus Torvalds and other kernel developers to be the “base standard” for the kernel. Some later additions to the POSIX specification have not been as useful as POSIX.1, and design issues often have to be resolved before Linux can make use of a standard. Linus describes one of the best examples of such a design decision:

Often there are standards that are too generic to be very useful as a guide for the kernel. The “pthreads” POSIX threads standard is one example: some people have tried to implement their kernel threading model according to it, and the standard simply is not very well suited to that.

Even in that case, I still wanted to support the standard; I just did not want to natively implement the standard as-is. So Linux clone was created: the infrastructure to do threading under Linux, on top of which you can implement pthreads and also other threading models.

It is difficult to list all of the other standards used today by Linux. TCP/IP, Ethernet and other formal and de facto standards form the basis for networking in Linux. The IBM PC is one of the best examples of a de facto standard. (This standard is now more formalized.) The PC allowed thousands, and then millions, of people to run Linux on a system typically used for Microsoft Windows. It is a lot easier to take over the world if you run on the standard hardware of the day. Just as significant, the GNU C compiler (without which there would be no Linux) is built on top of the K&R (Kernigan & Ritchie) and ANSI C standards.