I am not sure we are pursuing the right path for solving
the problem that Ulrich poses to us, that rm [a-c]* is dangerous.
I acknowledge the problem, as said before. But I do envision that
the same people that expects the 'rm' to only delete files starting with
lowercase characters, also would be very stunned when their "ls"
lists their Makefile's README's and .* in between the other files
in strict alphabetical order, and not in front.
So what these people really need is a shell using the POSIX locale, or an
enhanced locale that is compatible with the standard POSIX locale.
The solution to the problem should then be done as a part of the installation
setup of
the operating system, to set up the right locale for the shell,
including LC_COLLATE, which could be a specific setup parameter,
different from the rest of the locale categories.
Furthermore I do not think that a specific collating sequence
is a well defined cultural convention, it does not match normal
internationalization requirements - in laymans terms. The best
characterization of this "cultural convention" I can find is:
"What would you expect culturally with regular expressions
of character intervals, eg in 'rm' and 'grep'?"
Certainly not the level of cultural abstraction we normally
deal with in locales. I think the proposed APIs are built on
a bad concept/abstraction.
Keld
|