CAM is an adopted method in the SCSI-3 specification.
See under http://www.symbios.com/x3t10. In many ways
it's less than desirable from a simplicity point of
view, but because of some aspects of its complexity
it becomes easier to handle I/O scheduling and thread
based error recovery. It also supports a good passthru
mechanism for the 'diddly' devices (toasters).

I fought against CAM for years at Sun, but since then
I've come to realize that it has a number of advantages-
possibly enough to outweigh ill-informed opinions of
certain engineers who want to different things for the
sake of being different (sorta like the Novell and
Microsoft way of doing business). Possibly enough to
bring together the *BSD unix systems with at least a
reasonable base to start with instead of further
religious schisms- thus ensureing *BSD will become
a minor footnote to history. But I digress.

In any case- CAM would be too major a step to consider. I'd
suggest that one of following occur:

	a) Someone backports from Justin's CAM patches to
	FreeBSD-current into the -current NetBSD AIC driver.

	b) Someone looks at the current (5.0.15) Linux driver
	for the requisite similar changes and does that to NetBSD
	as well.