+ darwin*)
+ # Darwin 7.5 has nl_langinfo(CODESET), but it is useless:
+ # - It returns the empty string when LANG is set to a locale of the
+ # form ll_CC, although ll_CC/LC_CTYPE is a symlink to an UTF-8
+ # LC_CTYPE file.
+ # - The environment variables LANG, LC_CTYPE, LC_ALL are not set by
+ # the system; nl_langinfo(CODESET) returns "US-ASCII" in this case.
+ # - The documentation says:
+ # "... all code that calls BSD system routines should ensure
+ # that the const *char parameters of these routines are in UTF-8
+ # encoding. All BSD system functions expect their string
+ # parameters to be in UTF-8 encoding and nothing else."
+ # It also says
+ # "An additional caveat is that string parameters for files,
+ # paths, and other file-system entities must be in canonical
+ # UTF-8. In a canonical UTF-8 Unicode string, all decomposable
+ # characters are decomposed ..."
+ # but this is not true: You can pass non-decomposed UTF-8 strings
+ # to file system functions, and it is the OS which will convert
+ # them to decomposed UTF-8 before accessing the file system.
+ # - The Apple Terminal application displays UTF-8 by default.
+ # - However, other applications are free to use different encodings:
+ # - xterm uses ISO-8859-1 by default.
+ # - TextEdit uses MacRoman by default.
+ # We prefer UTF-8 over decomposed UTF-8-MAC because one should
+ # minimize the use of decomposed Unicode. Unfortunately, through the
+ # Darwin file system, decomposed UTF-8 strings are leaked into user
+ # space nevertheless.
+ echo "* UTF-8"
+ ;;