One of the most entertaining aspects of the rise of open source has been Microsoft's shifting response to it. At first, its line was simple dismissal: nobody used it. But as successive market share reports showed quite clearly that its adoption was on the rise, particularly among Microsoft's core corporate customers, this position became untenable.
So Microsoft moved on to technical criticism: yes, a few people were using it, but only for low-end tasks. The fact was, it insisted, GNU/Linux does not scale. And this was true - for a while. After all, the Linux kernel had begun as a bedroom hack on a PC with 8Mbyte Ram and a 386 processor, so it is hardly surprising that its capabilities were limited at first. But just as GNU/Linux's market share grew, so too did its technical capabilities.
Today, enterprise GNU/Linux users have a far wider range of hardware choice than those running Windows. IBM, for example, sells systems with two- to eight-way Power processors. SGI offers its Altix system with four- to 12-way Intel Itanium 2 processors, as well as a supercluster configuration. Alternatively, there are the well-established Beowulf clusters.
Beyond clusters lie grids, or, if you prefer big iron, you can run GNU/Linux on an IBM zSeries mainframe. And if you want seriously big iron, you could always wait for commercial versions of the 200-teraflop Blue Gene/L supercomputer.
When it became clear that the technical argument was shaky, Microsoft tried a different tack, focusing instead on intellectual property issues. In 2001, the company's chief executive Steve Ballmer said, " Linux is a cancer that attaches itself in an intellectual property sense to everything it touches." But as Microsoft's own research showed, attacks of this kind not only failed to impress customers, but were counter-productive.
Another strand in Microsoft's argument that customers should not adopt open source solutions, which is rather more plausible than the others, is that the free software movement is incapable of innovating. Or, to use the colourful phrase of senior Microsoft executive Jim Allchin, it is always "chasing tail lights".
This is a much more interesting issue than those concerning market share or technical capabilities, because it raises a question about the very nature of the open source development methodology: can even gifted hackers create major software advances, or are they doomed to merely coming up with neat ways of re-implementing other people's ideas?
The case of the most high-profile success of free software, GNU/Linux, certainly seems to lend credence to the idea that the best work is done by copying the efforts of others. But this conveniently forgets that both the GNU and Linux elements copied Unix for reasons that had nothing to do with lack of originality.
Richard Stallman chose to model his free operating system on Unix so that people who had written Unix programs could run them on his system, and Unix users could make the transition to a free operating system more easily. Linus Torvalds wrote his Unix-like kernel because he was studying Unix as part of his computer course but couldn't afford to buy a commercial version.
And even if GNU/Linux is unoriginal in a sense, there are plenty of free software programs that were innovative when they first appeared. These include the Berkeley TCP/IP stack, Bind and Sendmail, which helped kick-start the entire internet revolution; the world wide web, which took it into the mainstream; and Perl and Apache, which made e-commerce possible.
But if Allchin's comment is wrong historically, it does raise the further important question: is open source continuing to innovate today?