I'd hazard a guess that far fewer than 1% of Windows desktops have
any version of the .Net framework installed today. While WinXP has made some penetration in the home market, it is a rare company that has had the resources in this economy to replace a sizable fraction of its "fleet" of desktops in the last 6 years, and even less so for the past 3 years. For this reason most are still on Windows 2000, if not 95/98/NT4. Businesses don't typically replace the OS that came with their machines.
If they do start moving away from Windows 2000 to XP (instead of sitting on the fence hoping for Longhorn's release so they can sit on the fence waiting for SP 1), then I think we'll see
some version of the .Net Framework more widely deployed.
ASP.Net usage of beta .Net 2.0 is becoming widespread, mostly because of the "long" wait time since 1.1 came out and the need to chase the "newest thing" among the counter-Raymon Chen camp. This doesn't really impact the desktop one way or the other though.
The continuing problem with .Net's lack of desktop deployment has led to a dearth of sigificant desktop .Net code though. This is a vicious circle, because a ton of great applications might drive Framework deployment.
You may think 25 MB is nothing, but clearly somebody does. Why else is(are) the Framework(s) so rare on non-developer desktops?
This is a little old (4 months?) but
NET vs Java: The War Moves To The Desktop makes it clear worry over the desktop isn't solely a .Net issue. Also see the year-old babblefest at
Could .net fail?
What it comes down to is that there are growing pains, and we don't really know what killer event will have to occur before this sort of speculation is history. I think .Net is inevitable myself, if only through the sheer momentum of Microsoft's .Net push to date.