When someone is going to pay $3,000 for a Windows NT or Windows 2000 license, you can bet she is going to use it. If she pays $2 or even $50 or $150 for a Linux package, that doesn't necessarily mean it is going to get used -- or it could mean it was used 1,000 times by the organization. We just don't know how to count that.

Microsoft is very good at momentum marketing. It can turn reports of strong growth in revenues, software licenses shipments, or clients being supported into a message that Windows NT is becoming the de facto standard.

If you look at the overall Internet, there is a mix of operating systems.

For database servers, app servers, and Web servers, the tools are all there now, or have been announced and will be shortly.

Media reports often leave the impression that Windows NT is being adopted by organizations of all sizes for every conceivable mission and that organizations are abandoning their investments in other operating environments. However, when IDC shines the light of empirical research on Windows NT usage, a different view emerges.

Our estimation of how Windows 2000 will be adopted is fairly slow. While that trend is going on, we [also] see the growth in Web-oriented applications, and as they move in that direction, the tie to Windows 2000 is lessened.

It's a market that has been really difficult for us to get our arms around. We have no real way to track the number of copies of Linux that are installed. If a person downloads a copy from site, they could wind up installing Linux on one machine or a thousand machines. What we can keep track of is the money generated by commercial shipments.

In many cases, Windows NT has been brought in to work alongside the operating environments that were already in place in the mid-1990s -- supporting the organization's application portfolios rather than replacing them.