On Sun, May 20, 2012 at 11:36:35AM -0700, Linda Walsh wrote: > Anyway... so WHY does bash collate this way?
It doesn't. The operating system does. Bash just calls upon the C library's strcoll(3) routine. The results vary across operating systems, and even potentially across locale definitions within a given OS implementation. > Under what rules is bash > operating? I.e. justification? http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap07.html In particular, LC_COLLATE starts at http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap07.html#tag_07_03_02 > If other programs claim to have locale specific sorting & character > collation, should they be sorting the same way? Yes. ls(1) for example sorts filenames according to the collation order defined in the current locale (LC_ALL or LC_COLLATE or LANG). You should get the same sorting from both ls and *. For instance, on HP-UX 10.20, in the en_US.iso88591 locale: imadev:~$ mkdir /tmp/greg && cd "$_" imadev:/tmp/greg$ touch a A á Á à À â Â ä Ä b B imadev:/tmp/greg$ ls A a Á á À à Â â Ä ä B b imadev:/tmp/greg$ echo * A a Á á À à Â â Ä ä B b imadev:/tmp/greg$ LC_COLLATE=C ls A B a b À Á Â Ä à á â ä imadev:/tmp/greg$ LC_COLLATE=C; echo *; unset LC_COLLATE A B a b À Á Â Ä à á â ä Meanwhile, on Debian 6.0, in the en_US.iso88591 locale: arc3:~$ mkdir /tmp/greg && cd "$_" arc3:/tmp/greg$ touch a A á Á à À â Â ä Ä b B arc3:/tmp/greg$ echo * a A á Á à À â Â ä Ä b B As you can see, the two en_US.iso88591 implementations are not the same. See http://mywiki.wooledge.org/locale or myriad other resources (including the POSIX page linked earlier) for further explanations.
