> May I ask a different question:
> 
> But why do we need to split ports? The result is that we have to build
> the distribution N times, while 90% of that code is shared. Where 90% is:
> 
This does seem a bit wasteful of CPU cycles. I hadn't really though 
about it being a consequence of doing the port split.

> - the MI kernel code.
> - most of the CPU specific kernel code
> 
> It would be more wise, IMHO, to work on the installation process so that
> it installs the right kernel for the target machine, the right booter, and
> the right /etc/ttys (actually, if only Shark had wscons...)
> 
I've got the ttys problem sorted for the install kernels. I'm just 
waiting for some feedback from someone to make sure it works okay on a 
CATS machine. However it only deals with the installation /etc it does 
not patch up the /etc/ttys that gets installed.

> This way we could build all of the arm32-$X distributions in one run, using
> the fastest machine available (be it a Shark/EBSA/other-Strongarm with fast
> disk, or a fast Alpha/PentiumV with a cross toolchain).
> 
IIRC this is what Ben has in mind with trying to make arm26 and arm32
binary compatible for userland binaries. 'make build' would take an
insane amount of time to complete on an arm26 machine.

> Whoever wants to split ports even more should device a release building
> process that allows reusing the 90% objects already build on, say, some
> Strongarm build slave for building the, say, RiscPC kernel and install medi=
> a.
> 
Indeed. In the current world it is no problem building for SHARK or 
CATS on a RiscPC and vice versa.

Mike