Giga Research says that you should rule out Linux on the desktop until 2005. From what I've seen I think you should rule it out for the forseeable future. There is no evidence that Linux is gaining any serious "mind share" or market share in desktop computing. I've heard some decent things about Lindows, and there is some good software being written for Linux. But all of these efforts are targeted towards technology people. For a typical user, this technology is far outside of the mainstream. Longhorn will likely be available in 2005. For Linux to be able to compete on the desktop in 2005, it needs to be ready now. Does anyone really think that Linux will be able to compete with Windows on desktops that quickly?
Essentially, Linux is not positioned to succeed in the desktop computing market. The model for Linux development and distribution is not mature enough to support end users in the way that they're supported for Windows. Linux does not have a comparable selection of software for the average user. The way that Linux is positioned in the market must change significantly and resources in the Linux community must be organized around making Linux a reasonable alternative for end users. I can't imagine how this will happen, but until it does Linux is not a viable solution for desktop computing.