2005/12/7, Michael Stone <[EMAIL PROTECTED]>: > On Wed, Dec 07, 2005 at 07:54:59PM +0100, Tomasz Wegrzanowski wrote: > >http://mathworld.wolfram.com/LexicographicOrder.html > > That's a nice reference and all, but irrelevant. Or are you claiming that > sort somehow generates non-sorted data?
$ cat X a a a c ab c $ cat X | LC_COLLATE=C sort a a a c ab c $ cat X | LC_COLLATE=pl_PL.UTF-8 sort a a ab c a c This is not lexicographic. There's no consistent ordering between b and space. > The issue is that some symbols > simply have no meaning for certain kinds of sort (that is, they have > *no* relation such that A<B), which is different than saying that the > sort isn't lexicographic. You're arguing for a difference in the > interpretation of the symbols, not for a different kind of sort. (And > that *can not* be done within sort except by sorting based on ASCII > value (collate=C) and ignoring i18n, and eventually UTF, entirely.) It's not about ignoring i18n. Characters may be ordered in any locale-dependent way, but the lines should be sorted consistently or bad things will happen.
