Hi Tom :))

>>     Why must I install Python in order to compile the kernel? I don't
>> understand this. I think there are better alternatives, but kbuild
>> seems to be imposed any way.
>kbuild != CML2.

    Yes, sorry, just a mental shortcircuit ;))

>It all boils down to the current 'language' having no
>real definitive spec, and having 3+ incompatible parsers.

    Yes, I know and I think that is a good thing to have a good
configuration language, and it means having a good specification and
a good parser. Just I don't think that 6Mb-Python is a good way to
write a good parser. Well, I'm sure that I cannot do better (right
now), so I don't want to flame anyone with this, just want to show my
opinion (shared by many, although) and show the negative points of
having Python as a dependence.

>The spec for CML2 is out there, and there's even a CML2-in-C project.

    How advanced? Where is the spec, please?

>that project out and then convince Linus to include it.

    Hard job... Convincing Linus, I mean ;)))

>>     The kernel should depend just on the compiler and assembler, IMHO.
>The right tools for the right job.  C is good for the kernel.  Python is
>good at manipulating strings.

    Well, IMHO Python is good only in being big and doing things
slow, but... why the parser cannot be built over flex/bison?. That
way it can be 'pregenerated' and people won't need additional tools
to build the kernel.

    Raśl