I am writing this article in Montevideo, Uruguay, on the 68th anniversary of my birthday in 2018. It also happens to be the month that this magazine offers all 200+ issues on a DVD, dating back to October 2000, which brings back lots of memories for me.

These issues do not reach back to the beginning of the Linux kernel. No magazine does, because when Linus started the kernel project in 1991, it was "just for fun." Few (if any) people imagined that this fledgling project would significantly affect computing.

After all, in 1991, Microsoft ruled the desktop (with a relatively few "apples" thrown in) and proprietary systems (MVS, VMS, Unix, and others) ruled the data center. The idea that a university student would someday challenge these proprietary operating systems created by companies like IBM, Digital Equipment Corporation, Sun Microsystems, and others with a kernel written by a gang of "amateurs" was absurd. Microsoft was busy trying to advance their wunderkind WNT (Windows NT), and even the publisher O'Reilly seemed to have given up on Unix systems, producing more and more books about programming with Microsoft.

By technical standards, Linux was not the greatest of operating systems in the beginning. A single-CPU, 32-bit operating system with a filesystem that was relatively weak, the kernel took until 1994 to reach V1.0 status and be considered useful.

When symmetric multiprocessing (SMP) was started in version 2 of the kernel, it used the Big Kernel Lock (BKL) to manage the serialization of critical sections, whereas other operating systems like Digital Unix had fine-grain control of these sections.

Real time was also poor, as many of the kernel sections were not deterministic in the amount of time they would lock out interrupts.

A friend of mine who owned a firm that helped customers determine which commercial Unix systems would be useful to them called Linux a "toy" when I asked him about it in 1995.

Certainly another issue was a lack of applications. Hard-core Unix people could use GNU/Linux (the distributions started coming out in 1994) for their own work, but there were no "real" applications in the beginning.

However, a lot of Unix system administrators discovered that they could use GNU/Linux to repurpose older hardware, giving it new life as DNS servers, firewalls, thin clients, and other devices, saving them from having to buy additional hardware to do these tasks.

In 1994, another major event happened: the invention of the Beowulf-style supercomputer, what we call "HPC" today. Two men from NASA, Donald Becker and Dr. Thomas Sterling, conceived of and implemented this method of replacing far more expensive and complex supercomputers with many (often cast-off) PCs hooked together with "high-speed" networking.

This innovation created a flood of people who found they could now afford supercomputers made from "cheap PCs" to do a lot of the work for which Crays had been purchased.

It also created a nice market for the Linux/Alpha port, which started in 1994 and was finished nine months later; now, people could have Beowulf systems with 64-bit address spaces to process huge amounts of data. The Alpha port had another main feature in the history of Linux. It forced Linus Torvalds to make sure the kernel sources could support at least two architectures and, in fact, be set up to support N architectures.

Other applications started to show up, but probably one of the most significant was when the commercial databases started to arrive in 1998. Informix, Oracle, Sybase, and other closed-source databases joined the more "open" versions, and GNU/Linux could now be used in a database engine package. The database companies supporting their products on GNU/Linux made other companies sit up and take notice.

The next major step in the life of Linux came in the year 2000, with the advent of networked embedded systems.

Up until the 2000, most embedded systems were proprietary, closed-source systems written from scratch by companies for a specific market or platform. Memory was still very expensive, processors were still relatively slow, and, if the units were "networked," it was typically serial lines on a private network. No really sophisticated networking was needed.

In the late 1990s the need arose for these systems to join the Internet and speak TCP/IP. To fit a network stack onto these proprietary OS versions, and at the same time port all this code (some of which was in assembler language and very non-portable), would have been a lot of work.

Fortunately, there was already an operating system that worked on these chips, had the compilers needed, had the networking stacks, and was easily licensable and free of royalties. Almost overnight, Linux became the most-used operating system in new embedded designs.

Now, on the DVD in this magazine, you can read the rest of the story.