There is the old rule of thumb that says "never buy a car the first year it is introduced" and similarly, "don't buy an OS until after the first service pack ships," but waiting until 2008 to upgrade to an OS that ships in 2006 is ridiculous. (And the idea of waiting for the first service pack to ship is probably pretty dumb too.) Especially since we are now seeing glimpses of Vista and those glimpses do not indicate any overriding concerns about Vista's stability or performance. (I have not yet had the chance to look at Vista but I hope to soon.)
The idea that there are any IT shops still "downgrading" new hardware to Windows 98 is alarming -- to say the least. There may be sound reasons for downgrading to Windows XP. In fact, Windows 98 was the only version of Windows from which I kept repeatedly downgrading. I wanted to like Windows 98, I really did, but Windows 98 (all flavors -- 98/98se/Me) was always too unstable -- and just too damned slow (compared to Windows 95, compared to Windows NT 4, compared to Windows 2000 -- you name it.) The one great strength of Windows 98 that made me try it over and over again was Windows Update. Once Windows 2000 hit the streets -- I never looked back.
Any IT shop not running Windows Server 2000 or better in its machine room needs new management. Period. And any shop not running Windows XP on the bulk of its client workstations today is crippling itself. Sure, there are still little pockets of Windows NT Server being run here and there -- even in organizations with robust and security-conscious IT departments. The only way to shut these installations down is to implement security policies which cannot be supported under these legacy operating systems. The need for network security should trump all other concerns. In the enterprise, the stakes are simply too high.
There are certainly sound reasons for the enterprise to delay upgrading its OS of choice until the next lifecycle replacement of infrastructure hardware/software -- even if this means downgrading newly purchased workstations to maintain a uniform client software "build," but this does not justify permitting two or three OS releases to go by without upgrading. The day it shipped, Windows XP was more stable and performed better than Windows 2000. Today, Windows XP (with SP2) is a dramatically better product than it was when it shipped in 2001. I am quite certain that it will be the same with Vista. Simply put, any IT department which does not upgrade its operating systems as well as its hardware on a three-to-five year lifecycle is stifling the ability of the enterprise to function efficiently. Further, any IT department running Windows 95/98/9se/Me on any workstation connected to its network is placing its enterprise at grave risk.
Reading Colin Barker's article "Gartner: Wait 'til '08 for Vista" brings another thought to mind. In fact, this may be the point of Gartner's comments ...
The question of whether or not it is cost-effective to purchase upgrade licenses for Windows Vista needs to be considered. This is a different question than whether or not to continue to downgrade to Windows XP or before. As I said above, there may be sound reasons for downgrading to Windows XP (but probably not to Windows 2000) for some time after Vista ships.
At what point, though, does an IT shop decide to purchase upgrade licenses for a new OS? In an environment where a three-to-five year hardware lifecycle is the rule, and where the cost of the OS is bundled into the cost of the hardware, it is hard to justify spending as much as $100 per workstation (not to mention the man-hours involved) just to upgrade the OS. Even if Vista is a dramatically better product than XP, why buy an OS upgrade (and take the time to rebuild the machine) for a workstation that I am likely to replace next year? For many IT shops, the answer to the question above may be never.
Once Vista actually ships, IT shops should begin to evaluate it -- and plan for its ultimate adoption. That adoption will most likely be gradual as servers are retired and replaced with new servers sold with Vista licenses. Adoption on the client side may also be gradual as client workstations are retired or en masse once a critical number of new client workstations have Vista licenses. At that point, taking the opportunity to upgrade from a legacy software "build" to a Vista "build" makes a lot more sense. In either event, the upgrade will not be immediate.
In the end, whether an IT department buys a Windows XP workstation today or a Windows 2003 server, Gartner's advice (if not their rationale) is sound -- to wait until 2008, when you upgrade your hardware again, to upgrade to Vista. That is not to say that if you buy a new server a year from now (after Vista ships) that you should not consider using Vista. By all means, unless there is a good reason not to (such as legacy or mission-critical software incompatibility), you should move to Vista -- but this need not be en masse and it need not be at added expense. All Gartner may be saying is that there is no compelling reason to spend money that you were not going to spend already making sure that you have Vista in your shop as soon as it ships.