On Thu, 3 Sep 1998, Markus Kuhn wrote (quoting Clive Feather):
(9) All new functions to be re-entrant.
As noted before on this list: this means that timezones should be explicit parameters, and new functions should not have implicit dependence on the local timezone, though they may have such dependence on the locale. However the re-entrant functions from POSIX should maybe still be included despite not having explicit timezone parameters (ctime_r, localtime_r).
Anything else ?
Some more points that need to be covered: *) Allowing conforming implementations both with and without leapsecond tables. *) Retaining compatibility with systems on which time_t does not allow for leapseconds, as well as those on which it does, while still allowing those on which it doesn't allow for leapseconds (e.g. POSIX) to have leapsecond support and TAI<->UTC conversions through new interfaces. *) Allowing for real timezones (presumably through compiling timezone descriptions from a TZ string which could represent a POSIX timezone or an implementation extension beginning with ':'). -- Joseph S. Myers jsm28@cam.ac.uk