It looks as though the battle for the server market is being won by Linux. It was reported recently that most Wall Street trading operations are converting to the open source operating system.

This may not sound important but, for those like me who keep an eye on the markets, it is relevant. Some IT market sectors are strong technology validators and lead the way in adoption. Wall Street traders make up such a sector.

The irresistible rise of Linux has been interesting to watch. At Bloor Research we began tracking it in 1996, when a researcher we met from Edinburgh University remarked that Linux was spreading like a virus among students.

Software fads happen all the time, but those that spread through universities can be very influential because the students leave, become employees and get to make purchasing decisions. The rise of Apple and Digital were both assisted by this kind of propagation.

But Linux was different. There was an ideological edge to it; open source was both a technological trend and a quasi-political movement. In some ways this hindered it, and nobody marketed Linux properly for years. But it spread anyway.

Microsoft, Intel and the rest of the PC market were locked into a strategy of planned obsolescence. The main reason for the business purchase of PCs was the office applications, the so-called 'personal productivity' apps.

They had hit feature overload by the early 1990s and there was no meaningful way in which they could be improved to take advantage of increased power on the PC.

But Intel needed to move to the next generation of chips, and Microsoft helped by alternately upgrading Windows and the apps, changing the file formats and making the executables bigger. This was enough to drag the users along because PCs that were a few years old would soon begin to look out of date.

But what to do with the old PCs? One thing you could do was load Linux onto them. Linux was not subject to the eternal circle of greater power and greater software bloat. Old 486 PCs could become Linux servers of one type or another and, in many installations, they did.

It is impossible to measure the extent to which this has happened. In reality many companies simply do not know how much Linux they have running on old boxes on the corporate network.

With all other software you could count licence revenue and it would tell you roughly how many units were in use. This has never applied to Linux.

By 1999 there was enough Linux about for there to be a good skills base forming, and the students, mostly in the US, had begun to infect the software companies.

Suddenly, Linux versions of popular server software started to appear, so that major database, middleware and systems management products ran on Linux and, of course, Apache had taken the web by storm.

This was the moment when Linux became credible. The truth is that nobody really buys an operating system, or if they do, it happens very rarely.

You buy applications. In order for a platform to sell it has to have the applications. The software companies that write the applications do not like to support many platforms. Indeed, most of them support only one and will port the application to another system only if they can find a customer that will pay.

At the low end of the server market, Windows NT had become the developer's reference platform, effectively ousting Novell. And at the high end of the Unix market, the developer's primary platform was Sun Solaris.

In 1999 Linux stepped into both markets. Then in 2000 IBM announced a huge $1bn investment in Linux and threw its weight behind the Linux movement.

This was not driven by altruism but by the fact that IBM did not own either of the major developer platforms. It could not own Linux but it could become its primary commercial supporter.

Amazingly, IBM even ran a Linux advertising campaign: the only time I can think of when a commercial operation has advertised a product it did not and could not sell.

By early this year, Linux was spreading. It was being installed on old servers, and it was being bought for installation on new servers. Firms such as Egenera were building specialist hardware, which provided extremely powerful Linux servers.

In educational establishments huge networks and clusters of Linux servers with hundreds of machines were being assembled. Linux was also being used on personal digital assistants, thin clients and even smaller devices. It was even in use on consumer devices such as the TiVo.

Linux has won in the server market because it scales and is robust. It didn't win by slick marketing, because there wasn't much, apart from the cartoon picture of a penguin and a helping hand from IBM. It won by quality and viral proliferation.

This is bad news for Microsoft. Most of the firm's server products run only on the Windows server. If Windows becomes marginalised, then its server revenues will decline.

There is little chance of this happening quickly because of the wealth of software on Windows, but it could happen in time, and probably will. Sun is looking into a similar abyss.