A lot of systems have the C89 system call strftime, taking a buffer, a format string f (IMHO arbitrary bytes with ASCII format specifiers) and a broken-down time t.  It fills the buffer with a representation of t, based on the locale and f, in locale encoding.  The return value is the number of bytes written, not counting the trailing \0, or 0 if an error occurs.

== I am concerned about the following:
* the representation is not in UTF-8 and must be post-processed for displaying
* buffer filling always risks a buffer overflow and might also unnecessarily restrict what can be printed (maybe this is a subjective issue)
* there is no definite way to detect errors, because empty strings are valid representations, e.g. given %p in many locales.
* on Windows strftime must be used with care, see bug 404832.

== What I would love to see:
gchar *
g_strftime (const gchar *format, const struct tm *tm);

format - format string, in UTF-8, may contain normal text (and specifiers not available in the systems strftime, like %F on Windows?)
tm - broken-down time

Returns NULL on error or a UTF-8-encoded string represention of tm.