The problem is that sorting has two purposes: * sorting by scripts * sorting so the things are in the right order for humans
The first: * simply has to be lexicographical, otherwise too many things break * is much more common use of sort utility The second: * is less common * may use weird the locale-dependent sort order * but nothing fatal would happen if we sorted it lexicographically * and the default unicode ordering is wrong for half of the languages (like Polish) anyway So if we can't simply ignore those punctuation-ignoring broken locales, the sort should at least have a command line switch --human and --lexicographic or sth like that. Because using sort in scripts is much more common, and a lot of script may fatally break if sort is not lexicographic, I think --lexicographic should be the default. However, even if we want to keep the broken ordering default, at least providing a command line switch for doing it right would be an improvement.
