Vista will ultimately be a practical necessity for people dead-set on using Windows, simply because Microsoft is not working on advancing the hardware support of earlier releases of MS Windows. For instance, support for 64b processors, support for wider buses, and support for higher memory quantities will never appear in XP to the extent that it will in Vista.
Similarly, as application vendors start supporting Vista as their primary deployment platform, they'll stop supporting older MS Windows releases with newer application versions. There will come a time when, for instance, Adobe won't bother to make new versions of Photoshop compatible with older releases of MS Windows.
Because Microsoft will no longer support XP to the extent that it supports Vista, the world will move on, and you'll have to either upgrade your MS Windows version or move to a different OS platform entirely if you want to be able to make full use of newer hardware and software. The term that applies here is "planned obsolescence", such as using rubber timing belts that wear out in a few years and require replacing at great cost to the car owner. They could make things last longer, support software longer, ensure that timing belts don't wear out -- but they wouldn't make money on upgrades and replacements if they did that.
Of course, in a support-based revenue model such as is used by vendors like RedHat or SuSE, and with free/libre/open source software, revenue doesn't really factor into the software development process. As a result, with software like Linux, FreeBSD, and Plan9, you don't have to worry about planned obsolescence: you don't have to buy a new version to apply updates to what you've got. That's just one reason among many that I moved away from using MS Windows years ago.