Summary: Proprietary computer hardware gave way to commodity computer hardware. Is the next revolution going to be commodity operating systems?

Prelude: The Paleolithic Era

Once upon a time all computer companies were proprietary. They all produced their own hardware, incompatible with all other computer companies' hardware. They all created their own operating systems, incompatible with all other computer companies' operating systems. And the age of dinosaurs lasted long, and for many eons (in computer years).

Then a strange thing happened: The notion of an operating system as something apart from the hardware vendor. This seems to be a meme that spontaneously erupted out of nowhere, occuring to a college professor and a Bell Labs researcher at virtually the same time. The college professor, of course, was Gary Kildall, and his operating system was called CP/M. It ran on most "serious" microcomputers (i.e. computers that had disk drives) until IBM introduced their "PC". The biggest manufacturers of CP/M systems were companies whose names are little known today, companies like Cromemco, Altos, Kaypro, and Osborne.

Gary Kildall's company, Digital Research, did no hardware development. It was strictly a software company that specialized in improving CP/M and providing programming tools for CP/M. In the meantime, the two Steves bought a BASIC interpreter from a guy by the name of Bill Gates to go with their little Apple computer. That BASIC interpreter, with some disk-oriented extensions, became their first operating system -- but that way, inevitably, led back to proprietary operating systems, because Bill's BASIC had never been designed to be an operating system and thus every person who hacked it to make it an operating system did it differently. Compare Apple's hacked BASIC to Commodore's hacked BASIC to see what I mean.

The Bell Labs researcher... well, we'll never know exactly who it was. Was it Ken, or Dennis, or someone else? But the fact remains that Unix was invented specifically as an independent operating system. Unix, like CP/M, was tied to no one manufacturer's hardware. As a result, once AT&T was broken up and became free to sell Unix, Unix became for minicomputers and high-end servers and workstations what CP/M became for microcomputers.

The Mammals Arise

When IBM created their first PC, their PC division did not have a big enough budget to do things the normal IBM way (i.e., write a new proprietary operating system for their new proprietary hardware). Furthermore, it did not have a big enough budget to do the normal IBM thing of creating a full family of interface cards. So they did two things that were very unusual for IBM:

1. They bought an operating system from an outside vendor, and
2. They opened up the hardware specifications for their system.

The Last Hurrah of the Dinosaurs

Yet: Rather than learn the lessons of CP/M and MS-DOS, Apple chose to keep that wonderful OS specific to their own hardware. Thereby, ten years later, condemning themselves to irrelevancy.

Meanwhile, over at Commodore, total pandemonium reigned. They'd made (and lost) a fortune selling cheap 6502-based computers. They had warehouses full of returned Commodore 64 computers. Their successor to their quirky Commodore Pet, the B-128 with its companion SFD-1001 disk drive, was clearly not going to be competitive with the IBM PC or the new PC clones that were coming on the market, yet they had spent a ton of money filling warehouses full of those things too even though they never released it. Their engineers were working on a nifty little Unix-based widget in the back room, that was based on the Z-8000 processor and a neat little GUI that looked kind of Mac-ish, but there was no money to finish it.

So: First Jack Tramiel, the guy who'd gotten them into this mess, got booted. Then they spent money buying a struggling game machine company called Amiga Corporation, spent as little money as possible buying a DOS to run on top of the nifty multi-tasking kernel and GUI that had already been written by Amiga Corp. so that they could add disk drives to the thing, and went into the proprietary hardware and software business. In the end, we know what happened: They got creamed. They never had the money to both finish the OS and keep the hardware up to date. If they'd only had to do one or the other they may have been able to make it, but Commodore Business Machines (R.I.P.) was final proof that the days of proprietary hardware have ended.

The Mammals Evolve

Meanwhile, the OS manufacturer lucky enough to catch the wave of commodity hardware, Microsoft, rides high despite the fact that it was not until 1995 that they had a product that could compete on its own merits against MacOS or AmigaOS. Thus people run a proprietary but hardware-manufacturer-independent operating system on top of their commodity hardware.

Higher up the food chain roam the medium-iron Unix workstations and servers, running a hardware-manufacturer-independent operating system on top of proprietary hardware. Higher up still run the last of the dinosaurs, the "big iron", running proprietary operating systems on top of proprietary software. They have evolved lately, reinventing themselves as "enterprise data warehouses", filling those few needs where bigger really IS better. So what is the next paradigm shift to hit the computer world?

I believe it is quite clear. If the shift from proprietary hardware to commodity hardware was inevitable from the first day that Gary Kildall compiled his CP/M operating system, I believe that the next shift is similarly made inevitable by the existence of commodity hardware: what else could it be but commodity operating systems?

The Evolved Mammmal

1. Commodity hardware.

   As I mentioned, by far the strongest force leading to commodity operating systems is commodity hardware. People have become accustomed to multiple vendors to choose from. Few people want to go back to the era of "nobody ever got fired for buying IBM", where you basically paid what IBM told you to pay and that was that. Yet in the operating systems market they are still stuck using a proprietary operating system from a single company. They are still at the mercy of that one company.

   IS buyers want a single source for their hardware -- but they enjoy having a choice of possible vendors to select that single source from. This ensures that they get low prices, and because the hardware is commodity hardware, they can be assured that it runs their current operating systems and software.

   Microsoft insists that IS buyers want a single source for their OS. That is true -- but what they really want is to be able to select that single vendor from a choice of vendors, and to have the power to switch vendors if one vendor does not meet their needs. Currently they can do this in the server market, but for desktop machines, they're stuck with whatever will run their current applications -- which, realistically speaking, means Windows 98 or Windows NT Workstation. The contrast between this situation and the hardware situation is striking.

2. The End of OS Evolution

   This is basically tied to the increasing power of systems caused by the commoditization of hardware. Microcomputer operating systems have basically evolved to their limits. There are no major advances in operating systems theory left to be applied to microcomputer operating systems. All current major operating systems (except for MacOS, which isn't really "major" anymore, sorry) have a similar set of abilities: multi-tasking, virtual memory, task protection (well, almost, sorry Windows 98!), networking, inter-process communications, threading. All that is left is the task of adding additional layers of value on top of the operating system, such as a nice GUI interface, additional network protocols, additional applications, etc. Recent modifications leading to Windows 98, Linux 2.2, and Windows NT 5.0 tend to emphasize that -- all focus on adding new protocols, updated drivers, and other such "widget frosting" as compared with adding additional functionality to the base OS kernel.

   Therefore the primary reason for proprietary operating systems -- that they created new value by supporting new features of microcomputer hardware -- is in the process of disappearing. The next new feature of microcomputer hardware likely to be added is 64-bit addressing, which has little impact upon OS design.

3. Economics

   Commodity hardware did not win out because it was better. It won out because it was economical. Commodity operating systems software benefits from that same advantage. Due to the economics of scale, it will eventually be cheaper for a vendor to add commodity components together and then toss on some "widget frosting" than it is for a vendor to write an operating system from scratch. An example is Red Hat Software, a small startup in Raleigh, North Carolina. Despite total sales of perhaps $10,000,000 per year, out of which they have financed the development of their operating system, they have put out a full-featured operating system by taking commodity components and putting them together with their own "widget frosting" (improved installation, nice manual, etc.). The full development cost to create such an operating system as a proprietary product would be close to $10,000,000,000 (ten billion dollars). Even if the commodity components being used by Red Hat Software did not exist, eventually a commodity OS component industry would have arisen for the same reason that a commodity OS component industry arose in the PC hardware market -- because commodity components, by adding flexibility and benefitting from economics of scale, are simply more economical in the long run.

4. Long-term credibility The commodity hardware market has long-term credibility. That is, you know that you can wake up tomorrow and the commodity hardware market has not disappeared off the face of the globe. Sure, individual companies may come and go. Micronics may go from being the industry leader in motherboards to being bankrupt and bought up by Diamond, Hayes may go from being industry leader in modems to declaring Chapter 11 (twice). But: You can rest assured that your investment will not evaporate overnight with the company that created it. Micronics gone? No problem. Just buy an Asus or Tyan motherboard.

   A commodity operating systems market benefits from that same long-term credibility. When no one company or person owns the basic underpinnings of an operating system, when there are multiple competing vendors selling the operating system but (as with commodity hardware) you have a guarantee that your programs will run on all vendor's flavor (and any vendor who does not provide that guarantee will be out of business quickly!), then that operating system has long-term credibility. You can base your business upon it without worrying that the company that produces it will go out of business, discontinue it, or raise license fees to the point that they adversely impact your business.

5. The Internet

   The rise of the Internet has had the unexpected side-effect of allowing collaborative efforts to happen on a global scale. This has allowed projects such as FreeBSD and Linux to happen, and revived the once-moribund GNU Project (which had pretty much reach the limits of what could be done with the previously-existing UUCP-based network). In addition, it adds a large market for a commodity operating system: as the basis for networked devices which do not present an operating system face to the outside world. Do you know what operating system your ISP's web server is running on?

6. Linux, xBSD, etc.

   These at the very least provide an example of what a commoditized operating system would look like: something that is not particularly revolutionary, but which creates a base upon which your own systems can be built. I will note that I successfully used Linux as the base for an educational administration solution, allowing us to underbid proprietary solutions while providing equivalent function. This was possible because we sold a complete solution -- all that the secretaries knew was that they had this "PAMS" thing in their office that presented our custom software on the screen. None of them knew that they were using Linux.

7. The final straw: The WINE Project

   The WINE project is coming increasingly close to being able to run all 32-bit Windows applications. What this will do is immediately commoditize the operating systems market by allowing buyers to select from whichever vendor they like most and rest assured that it'll still run all of the software they've invested so heavily in.

   The Future of Mammals

   As mentioned, commodity hardware has killed the proprietary-hardware personal computer (and I count Apple as killed, even if they don't know it yet). Hardware vendors, rather than inventing their own proprietary platforms, mix and match commodity hardware and add their own value-added items to it in order to differentiate themselves from their competitors. So what will life be like in a commodity operating system world?

   The first thing to think about is this: A commodity operating system will have a common base, a set of commodity components that can mix-and-matched on top of it, and then OS vendors, rather than inventing their own proprietary OS, will mix-and-match commodity software components and add their own value-added items to it in order to differentiate themselves from their competitors.

   On top of this commodity operating system runs a variety of proprietary applications programs. That is, base framework is commoditized (both the hardware and the operating system), but a commodity operating system in no way puts an end to the commercial software world. Rather, it simply means a shifting of resources towards adding value rather than stifling competition.

   If this is already looking like the situation in the Linux world, you're starting to get the picture. In the Linux world, all OS vendors start with a common base (the Linux kernel and base GNU utilities). Then they mix-and-match commodity software components and add their own value-added items to their particular release in order to differentiate themselves from their competitors.

   Does this mean that Linux, then, will be the commodity operating system that takes over the personal computer market? Well, not necessarily. For example, Microsoft might decide to commoditize their operating systems by giving them away, source code and all. And cows might fly too. Hey, it could happen.

   Anyhow, needless to say, it does appear that Linux is poised to be the commodity operating system that brings to an end the sabre-tooth tigers and wooly mammoths of the software industry -- the proprietary operating systems vendors. Linux has the momentum and the technical ability (due to its origins as a clone of Unix). Keep watching. Things are about to get interesting.

   Evolutionary Dead Ends

   Or: The Fate of the Sabre Tooth Tiger

   What, then, becomes of proprietary operating systems vendors such as Be and Microsoft?

   I can see two possibilities.

   1. The most likely: They leave the operating systems business. Or:
   2. They re-invent themselves as vendors of commodity operating systems (can anybody say "Microsoft Linux?"), the way that IBM, after their disasterous flirt with proprietary hardware (the MCA-based PS/2), reinvented themselves as a vendor of commodity hardware.
   Either way, I believe that the next ten years will be exciting. It took ten years for commodity hardware to largely banish proprietary hardware from the consumer market (iMac nonwithstanding -- even the iMac is too little, too late, and represents the last gasp of an evolutionary dead end). It also took those same ten years for hardware-manufacturer-independent operating systems to banish hardware-manufacturer-proprietary operating systems from the consumer market. Ten years from now, I suspect that the same will have happened to proprietary operating systems: they, too, will be a swiftly vanishing remnant of an earlier era.