Warnings about rcs_id strings: let's settle this

[Joe Buck]

There has been an explosion of threads about the rcs_id warnings issue,
and a whole lot of people are getting sidetracked on unrelated issues
(like the semantics of "volatile", because of the suggestion that
people declare their RCS strings volatile).  These threads have buried
the fact that we have a reasonable solution.

Kean Johnston has posted a patch that splits the warning about static
variables into a special case.  It is in 

http://gcc.gnu.org/ml/gcc-patches/2003-05/msg00157.html

The reason I feel that we must do this is that otherwise we are asking
many, many software development groups to change every single file,
given the software management practices we know them to have, and it's
easy.

There are some details to be worked out; some of them are mentioned
in the discussion of the patch.  But I feel strongly that we need to
get this into 3.3, so we should try to settle this soon.