Understanding Unix

How many roads must an operating system walk down before we call it a keeper? This question has plagued the UNIX community for over 30 years. When the First Edition of UNIX System Administration Handbook was released in 1989, it seemed like UNIX then 20 years old had come of age. So much so that we could unabashedly discuss "touchy" issues such as how to wire-wrap the backplane of the then-popular VAX in public (a secret ceremony previously known only to Digital field engineers).

Now, few claim to have ever seen a mythical VAX. How times have changed! Then, when the second Edition of UNIX System Administration handbook was published in 1995, it appeared that the operating system had been successfully and completely accepted commercially. That edition discusses six of the commercially available UNIX platforms that were available at the time.

Intel-based variants were popping up all over, pairing low-priced hardware with high-functionality and high-reliability. Only happy days could be seen ahead: UNIX would flourish as the great enabler of the Internet and everyone's SUV would sport a "UNIX: Live Free or Die" bumper sticker on the back.

Suddenly, a very dark era descended on UNIX -- in 1996 we began wondering if it was destined to be completely replaced by operating systems from the great state of Washington.

By 1997, I was convinced that the second edition would be the last book we would write on UNIX System Administration, perhaps leading us to a life of pondering better ways to edit the Windows Registry.

But then the tides turned, slowly at first, but unmistakably. Demand for copies of UNIX System Administration Handbook exceeded expectations. E-commerce platform vendors such as Broadvision, ATG, Vignette, and BEA pushed their products on UNIX variants. Large databases in Oracle on UNIX platforms became accepted. Linux became popular, and though technically "not UNIX," provides the same basic OS abstractions (for my purposes, I consider it a UNIX variant). And questions began surfacing about Microsoft's business practices.

By 1998, it was clear that UNIX was back as a contender. But why? Was it just more market politics? Was is that Microsoft's solutions had some fundamental flaw? Was it that the folks who used and loved UNIX in their 20's were now moving into positions higher up in organizations as IS managers and technical leads, forcing their radical views down the throats of others? As we set out to write the third edition of UNIX System Administration Handbook, I struggled over these questions and others.

I think the answer to UNIX's revival is much more abstract. It's about evolution. The environment that UNIX has been in over the last 30 years has forced it to evolve, in a process akin to natural selection in biology. It is not an operating system that has been driven by a master marketing plan, or by some genius that is allowed to be late to board meetings, or by a multi-national committee. Instead, lots of variants (and sub-variants) have tried to survive in the marketplace. Some variants have died, never to return to this planet. Some have mutated into something even more bizarre. Some have evolved gracefully into beautiful, fully articulated rock stars. And finally, some of the genetic engineers have tossed everything into the laboratory blender and produced a high-featured hybrid. These folks, working on Linux and FreeBSD, are accelerating the evolution of UNIX.

Granted, there are still a few things in the UNIX gene pool that have survived but probably shouldn't have. Some good examples of this include both popular printing systems that are completely obtuse and inflexible, and Solaris' ridiculous tty/getty "service" management scheme.

An amusing, and maybe even insightful (at least in the trivial-pursuit sense) realization of the practical evolution of UNIX can be found in the goofy cover cartoons of the various editions of UNIX System Administration Handbook. Here, I'll attempt to point out the "highlights" in this regard, comparing the 1st Edition cartoon (circa 1989) to the 3rd Edition cartoon (2000).

The 1st Edition cartoon features a crude, hand-drawn rendition of what appears to be a car driving off a cliff. To the left is a sign "UNIX System Administration Made Difficult" which was a working title for the original manuscript (very unpopular with the publisher). In those days, UNIX administration was difficult, modern GUI tools hadn't been invented and UNIX system administrators had to survive an environment where they had to manage everything from glycol-based cooling systems, to processor board-swapping, to network protocol stacks that didn't quite work together.

At the bottom of the picture were the personifications of the popular daemons of the time. These included Biff, named after a (real life) dog, was the daemon that announced when the mail arrives. Cron, a guy with a watch, started processes that needed to run at regular time intervals. Fingerd, the finger daemon, let you finger information about people (we're still a little surprised this particular illustration got by the publisher). And Whod, the Owl, advertised who was logged into the system.

In contrast, the 3rd Edition cartoon had a much more polished, "evolved" appearance. In the upper left hand corner was a graveyard with a tombstone for Biff. It indicated the passing of the real life dog Biff, as well as the fact that few users were now using UNIX as their mail-reading platform. By this time UNIX had become the "mail server" platform which speaks IMAP or POP to (typically) PC-based mail reading clients. UNIX still underlies a large percentage of mail servers due to its scalability, network and security features. What has evolved is the way UNIX is now applied.

The graveyard in the 3rd edition cartoon also hosted UUCP, which some call "the first Internet." UUCP was a store-and-forward file and email transfer system designed to work over dialup modems. Most people today haven't heard of it, since it has been almost completely replaced by TCP/IP networking (except in extremely remote areas of the world). The OSI protocol suite, once thought to be the next big wave in networking, died a brutal death in the hands of the market.

USL, the AT&T-then-Novell attempt at "UNIX development by a board of directors," is also long gone. Curiously, UNIX development at CSRG (the Berkeley guys) is dead as well. This type of purely humanitarian development is now in the hands of the FreeBSD and Linux folks. In some sense, the grandparents of UNIX are gone, but the seeds they left behind have spawned dozens of other efforts to develop an operating system using the same philosophy and similar approach. The evolution of UNIX is in new hands, and no one is bitter about it.

The third edition cartoon also pictured a different set of daemons. These included the AMD mount-monkey, a representation of UNIX's ability to play in a highly networked file-sharing environment. There was also the world-wide web spider of httpd (handcuffed, for security), representing UNIX's adaptation to (and fueling of) the Internet and the web. Without the network building blocks and research results that UNIX provided, the Internet would not exist today. Finally, there was Doctor Snmpd (the management agent daemon), a sign that UNIX had evolved to an area where predictability, manageability, and monitorability was essential to survival.

Finally, on top of the car about to drive into the cliff, there was an antenna. By this time, UNIX was evolving into the wireless era as well, with devices such as wireless tablets designed specifically to run Linux. And, in the passenger seat, as shown in the illustration on this page, there was a penguin, symbolizing Linux.

My hypothesis for the future is this: True innovations (true evolution) in the operating system marketplace really happen in an environment where cross-pollination and natural selection is allowed to occur freely. UNIX has proven its ability to evolve, despite the marketing budgets of others. As a result, UNIX (and "not UNIXes" like Linux) is here to stay as a primary place where first order evolution takes place. It's where cutting-edge features can be deployed on a stable platform for situations where moving fast but mitigating risk matters. Hmmm, eMarkets, here we come!