Re: locale specific ordering in EN_US -- why is a<A<b<B<y<Y<z<Z?

[Chet Ramey]

On 6/27/13 10:54 AM, Paolo Bonzini wrote:

> 
> Fixing gnulib ensures that, on non-GNU systems, everyone will get the
> same result from gnulib users (BTW, note that grep's fix for the problem
> did not introduce a particular policy: instead, it just relied on
> gnulib's fix).  It's the best you can do on non-GNU systems.
> 
> But that's not enough on GNU systems, because GNU systems will always
> have regex in glibc and many programs will use the regex that is in
> glibc.  On such a system, modifying glibc is the only way to ensure that
> everyone will get the same result, either a<A<b or a<b<A.

You're sort of working around the edges of the problem.  It's bigger than
you describe here.  Each program that does matching has to use a matcher
that implements the desired behavior (and first you have to define that
behavior).  That means that you have to define the behavior, publish it,
get each program to implement it, and make sure that it persists.  Arnold
has done a good job of this so far.

Now, I haven't even mentioned the issue of script portability.  That means
the problem of something like 'ls [a-e]' being implemented consistently by
different shells that understand Posix shell syntax.  That's a whole
different level of hurt.

Chet
-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, ITS, CWRU    chet@case.edu    http://cnswww.cns.cwru.edu/~chet/