Just an opinion :-)

There are two build environments:

	o	native

	o	cross
		(from NetBSD or another system)

It's reasonable for a person installing a basic system to expect a
native rebuild to work out-of-the-box (which suggests that the native
build process was tested before product ship :-)

Separate to this, a developer may choose to use a cross compliation
environment (not necessarily NetBSD) as part of the devlopment process. 
When making such a decision, the developer would factor in a number of
issues including the reliability of the build machine and the buggyness
of the chosen cross compiler.

If NetBSD are going to seriously use cross compilation then there will
need to be:

	o	an identified set of reference NetBSD
		build systems (one BE, one LE) that
		are supported.  If you don't there is the
		potential for 26*26 different toolchains.. :-)

	o	an identifed compiler(s) used for cross
		compiling for each target. It might be the
		native compiler, it might be based on a
		recent GNU release.

Anyway, my cross compile wish list is to see:

	o	Eliminate the technical barriers that stop
		developers using any ``UNIX like environment''
		when building NetBSD.  (ex, try building
		an 1.4.1/i386 boot disk without NetBSD).

		This is just a return to the good old days
		when the kernel hacker used what ever system
		was available to do kernel development.

	o	(personal hobby horse)
		Re-do (cleanroom) any significant work done on
		the GNU toolchain so that it can be merged
		back in to the main GNU sources and base the
		cross compilers on that.

		In all likelyhood new development will use the
		latest GNU tools and not the NetBSD tools.

		The more stuff that is merged into the GNU
		source tree the less work (in the long term)
		that is needed to build/maintain various toolchains.

enjoy,
	Andrew