Unix, and the variety of operating systems (OSs) that closely resemble it, has an unusual background. Until a few years ago it was a relatively niche OS, regarded by most as 'too hard' for general use.

Its main use was in server and high-end workstation operations, and Unix implementations such as HP-UX, SunOS/Solaris and IRIX were made specifically to fit their manufacturers' hardware (Hewlett Packard (HP), Sun Microsystems and SGI respectively).

Then Linux arrived. Suddenly people were installing a Unix-like OS on their home PCs. University students were running Linux-based web servers from their halls of residence. Forward-looking divisions of otherwise backward companies were pulling old 486 PCs out of cupboards, running up Apache and experimenting with cool new stuff like the web and internet email.

Getting to the desktop
So today, Unix (a name we'll use generically throughout this article, by the way, to mean anything that looks like Unix) is growing fast in popularity and versatility. But what is the future for the OS?

Will commercial implementations give way to free ones? Are the days of paid-for Unix numbered? Will it even make it to the desktop, the last remaining Microsoft-heavy area?

In the past, the choice of OS was made for us in most cases. If you bought a Sun machine, it ran Solaris. If you bought an SGI box, it ran IRIX. If you bought a Mac, it ran MacOS (Apple's A/UX flavour never really took off). If you had an Intel box, it probably ran either NetWare or NT.

Then in the early 1990s, two things happened. First, free Unix implementations such as Linux blinked into existence; and second, the idea of cross-platform Unix started to take off. So today you can run Solaris on a Sparc system or your Intel-based server or desktop. Likewise, you can run Linux on your Mac, PC or Sparc machine. So the Hobson's choice of OS is no more - we can choose not only the hardware but the OS as well.

We'll look initially at server systems, as the situation is fairly stable and is unlikely to change significantly in the near future. The first choice today is often between Unix and 'not Unix'. Microsoft's Windows NT/2000 product, though traditionally a bit wobbly, gets more stable as time goes by (and service packs multiply).

Unix specialist Chris Boross said: "Now that Microsoft has a product that will sit on a server and serve a web page without the need to reboot the machine every few hours, the company seems to be making headway in the internet market. It's just so easy to roll out a network infrastructure or web serving farm, that some geeks and network engineers can't resist."

NT/Win2000 will continue to have a significant presence in the market because, whatever anyone says, it's fairly stable and simple to use. Unix is still regarded as 'difficult' by many, which is fair when you remembers that NT/Win2000 looks just like the Windows GUIs people are used to on their Win9x desktops.

That said, the Windows server OS family will continue to dominate only the corporate server market, as Unix is already significantly more popular for internet applications than Windows. The usual question for internet-related servers, then, is 'which Unix?'.

At the high end, manufacturers of the big, scary systems - HP, Sun and SGI - predominate, and will continue to do so, for hardware at any rate.

According to Greg Cope, a freelance Unix and networking consultant: "Unix is now dominant in the mid- and high-end areas. In the low server end Microsoft dominates with NT4, but this is increasingly under threat from Unix. It would be quite reasonable to expect Microsoft to be squeezed in this area as customers are unwilling to upgrade older hardware and/or install Win2000, which now represents a large proportion of a low-end server's cost when compared to a similar free open source system."

For the really high-end applications, the manufacturers' own versions of Unix will be an almost automatic choice, even though OSs such as Linux are in theory capable of running quite happily on high-end machines.

The reason is simply that there's a lot to be said for a commercial OS that has had thousands of hours of development and testing, not to mention optimisation for relatively rare machines such as the real high-end Sun or HP servers. As you come down the range, though, there will be an increasing temptation to run Linux on systems that would have previously been vendor-specific Unix machines.

Platform support
For low-end and mid-range internet servers, however, Linux and, to a smaller extent, FreeBSD predominate, and will continue to do so. Even when these systems were freely downloadable, but completely lacking in formal support, their sheer stability was enough to persuade companies to adopt them for their web platforms.

Now that we have companies offering commercially-oriented distributions and Linux/FreeBSD-based after-sales support - and even applications providers like Oracle and Informix offering Linux versions of their products - these free Unix incarnations are even more suitable for internet-related servers.

Most of us know that Windows predominates on the desktop, in both the OS and the applications. In the past two years, however, Linux has caught up significantly with the Windows world.

While it is fair to say that it's still not as easy to install Linux as it is Windows, this perception is almost entirely due to familiarity - Unix looks different and doesn't ask the same questions as the Windows installer.

The thing with the Windows versus Linux on the desktop debate is application support. It can't be denied that desktop applications are far more common under Windows than they are under Linux, and that people who have been used to the various versions of Microsoft Office over the years can figure out how to fly new versions with little difficulty.

But is this about to end? Mark Ray, of web development company Luminas, thinks so. "Sun and HP are probably losing the low-end to Linux," he said. "How far that goes depends on what tactics Sun and HP use to block it, which nearly always results in shooting oneself in the foot, or adopt it, which sometimes helps and sometimes hinders."

If you take the time to look, there are office-type applications pre-installed with many Linux distributions. Even if you don't like these, packages such as Sun's StarOffice (which has just been released in an open source incarnation) provides the desktop user with word processing, spreadsheets, drawing applications, email, presentations, web page building, desktop databases and everything else most of us have on our desktops.

So for those who do run-of-the-mill office stuff on their desktop computers, Linux does make a suitable platform on which to do it. Not only this, but, unlike Windows 9x, the OS is built around an inherently secure file permission and authentication system, so defining who can see what on whose machine is a doddle.

As for support, while Windows is fairly easy - if it crashes you reboot the machine and the problem usually goes away - any Unix-like system can be supported properly, even if it does need more highly-trained staff.

If something goes wrong with Unix, you can fiddle with the broken bit without affecting the rest of the system. (Note, for instance, that Unix doesn't have a blue screen of death because it does memory protection properly.)

While Windows NT Workstation is far nicer to support than Windows 9x, it's still not quite as stable or manageable as Unix, though admittedly you have to be less of an expert to do it. And if you really do want to run Windows programs, that's fine - you can do it under an emulator with Unix.

Windows will, for the foreseeable future, continue to be the most popular OS on the desktop. Just as NetWare lost its near monopoly on the server to NT, though, so Windows will lose its near monopoly on the desktop to Linux. It will take a while, and it's hard to imagine Linux seeing much more popularity on the desktop in three years' time than MacOS has today, but when you realise that this adds up to a few million desktops, it's not insignificant.

For those who want to do heavy-duty number crunching or graphics manipulation on the desktop - the traditional place for Sun, SGI and HP workstations - the OS will tend towards Linux rather than the manufacturers' own Unix flavours.

With Linux, you can tweak to your heart's content rather than being constrained by a system with a kernel you can fiddle with but which you can't rewrite bits of. Think about it: how many companies with Sun workstations running Solaris use Sun's C compiler, or Sun's mail software? Open source is the order of the day - people tend to use the GNU C compiler, the BIND nameserver, the Postfix/Exim/QMail mail packages, and so on.

Even the manufacturers, such as Sun with StarOffice, are realising that open source is the way forward. Cope puts it succinctly. "Linux, FreeBSD and OpenBSD will continue to flourish due to their openness, price, quality and attitude," he said.

So will the advent of new hardware such as the upcoming IA-64 from Intel have a significant influence on Unix's future? In short, not really. There's a deal of interest from many manufacturers in the IA-64 - HP announced its intention to adopt it some years ago, for instance - but ultimately there's more to an architecture than a processor.

New processors mean faster and better computing, but this is really a hardware issue, not particularly an OS one. Since, as we've seen, there is now a choice of OSs for a particular architecture (and IA-64, once it is established, will be no exception), when you buy a new, screamingly fast IA-64-based machine, you'll have a choice of IA-64-optimised Unix and Windows implementations to run on it.

Smooth transition
So while all this evolution of hardware and OSs is happening, is it possible to make the transition seamless and not interrupt work too much? It is because with Linux it doesn't matter what servers are being used, or that during the replacement process you have a heterogeneous network.

Out of the box, Linux can talk to NT, NetWare, Unix and Mac fileshares, and can, with little effort, act as a fileserver for any or all of those as well. Likewise, commercial versions of Unix can easily be made to talk to NT fileshares and printers, and NetWare can integrate with both NT and Unix. One bonus of the OS competition of recent years is that heterogeneous networking has become a reality.

The future for Unix is rosy or, more accurately, the future for free Unix is. Commercial Unix will continue to dominate the high-end market because it's unreasonable to expect a free implementation to outperform commercial offerings on rare, high-powered, highly tuneable equipment.

Unix, and Linux in particular, will continue as the dominant offering on low-end and mid-range internet-related servers, as well as taking a larger share (though far from a dominant one) of internal corporate file and print serving.

On the desktop, Linux will become a significant player, as it is rapidly becoming simple enough and sufficiently supported application-wise for this to be realistic, although Microsoft's current dominance is unlikely to be dented greatly in the next three years.

The final word on the subject of Unix's overall dominance goes to Cope: "As much as I would like to see the dominance of Unix, I would not want it. In evolutionary terms it is always best to have a competitive environment, where competition is healthy.

"Dominance leads to stagnation, which is ultimately bad. Hence, it would be best if Unix became a bit more popular, but not much. I for one would miss batting the Microsoft supporters."