How do we implement the conversion functions? The approach recently added to the CVS version of GHC is to use the native libraries, which requires the user to set the locale appropriately. More generally, should these functions be locale-dependent at all?

No they shouldn't be locale-dependent, but unfortunately that's what the C library gives us at the moment. wchar_t should ideally represent Unicode, but sadly it doesn't (with glibc) unless you set the locale to something/anything other than "C" or "POSIX". The situation is worse on Solaris: I had to set the locale to en_US.UTF-8 before I got correct results. I haven't tried FreeBSD yet. Windows gives the correct results without having to muck around with locales (but not if you use cygwin). So GHC's solution is patchy at the moment, but I hope the situation will get better in the future as more OSs jump on the wchar_t==Unicode bandwagon. Cheers, Simon