Hardware needs more than a driver for proper support

Seconded. Having a basic working driver is a good first step, but often users expect more from
their hardware. If we only look at driver status as a binary distinction between driver exists
vs. driver missing, then from a user's point of view every device works in principle but not
well in practice. 

This is still a huge step forward from the situation some years ago when a lot of hardware
actually didn't work at all with Linux. If the situation with most classes of hardware is as
good as Greg makes it sound, then we need to raise the bar higher and not just talk about
missing drivers.

As a personal story, I have a Dell X300 laptop with a memory card reader that barely works
with the out-of-tree sdricoh_cs driver. IO is much slower than in Windows because the driver
only does polling, no interrupts or DMA. I can't use the reader in practice to transfer
pictures from my camera, it would take forever. I don't expect the situation to improve since
no specs are available and the hardware is likely to be obsolete by now (so this is in Greg's
"hard" category #2). But on a checklist, the card reader might be considered supported.

Another example is the Intel video on the same laptop. Ubuntu Gutsy was the first distro that
actually allowed me to make full use of an external 1280x1024 TFT display in a dual-head
setup, thanks to the new RandR support in Xorg. Previous versions of Xorg either crashed
outright or, at best, forced me to use ugly XGA resolution on the TFT. I still (as of Hardy
beta) can't switch between laptop-only and dual-head modes without logging out of X (and often
having to use a messed-up console to restart GDM). In WinXP display hotplugging has worked
(almost) perfectly for at least two years.