It has the advantage of giving people an easy "try before you blow away
your system" install, which, as you point out, will not perform to 100% of
the capability of the hardware, but the install should be as simple as
unpacking a SIT/SEA file, and rebooting with no fuss.

We should, of course, measure the performance difference, so that we can
tell the user how much additional performance will be gained by doing a
completely "native" install on his hardware. As for the i386 world, why
shouldn't we do the same there? If we can reduce the amount of pain people
go through to try NetBSD, I bet we'd get a lot more converts.

The nice thing about doing this for the iMac is that, the most recent
"special" one aside, they have a known, relatively fixed hardware
configuration (the main thing people are likely to change is the amount of
RAM in it), which we can use to our advantage in such an install by
preconfiguring the system to come up in X with some relatively nice window
manager.

Of course, until we get that prerequisite HFS/HFS+ support, this is simply
an idea. Just a SMOP.

	Erik <fair@clock.org>