Re: [kbuild-devel] Converting the 2.5 kernel to kbuild 2.5

[Michael Elizabeth Chastain <mec@shout.net>]

I suppose a polite title for me would be "maintainer emiritus"
of make config, make menuconfig, and make xconfig.

I'm in favor of abandoning the current tools because:

  It's 3x maintenance to have 3 parsers for the same language.

  It's difficult to do good syntax checking in scripts/Configure and
  menuconfig.

  menuconfig in particular is too ugly to live.

  A company which considers Linux its #1 enemy may own the copyright to
  "scripts/Configure".  I don't know what kind of marketing or legal
  play they could make, but it would surely be hostile to Linux.

I'm in favor of CML2 in particular because:

  ESR has designed a clean theory, which the configuration process really
  needs after ten years of ad hoc extensions.

  ESR has done a lot of grunt work to turn a particular idea into a
  viable implementation.  It's hard to get that work done.

As far as the Python issue goes, I believe that the kernel documentation
just needs to state clearly what tools (and what versions) are needed
to build a kernel.  If other people prefer a C implementation, then
CML2 (the language) is amenable to a C implementation, so they can
write one.

As far as CML2 versus an mconfig-based solution, I am tilted towards CML2,
as it is simply a better language.  I would be happy with either choice
if Linus made one of those choices.  I would be unhappy if 2.6/3.0
continued to ship with Configure/menuconfig/xconfig.

Michael Elizabeth Chastain
<mailto:mec@shout.net>
"love without fear"