Inconsistancy between mktime and system time accross leapsecond

[Garrett Wollman]

<<On Sat, 06 Aug 2005 18:28:32 -0600 (MDT), "M. Warner Losh" <imp at bsdimp.com> said:

> So are we therefore precluded from fixing this obvious bug?  This is
> stupid and wrong.

Committee membership is open to all.  The 1003.1 standard is up for
review next year; you are welcome to try to persuade the sponsors
(IEEE, ISO, Open Group) that the current situation is unacceptable.
I did my part in 2001.

The reason this discussion always gets ratholed is that you'll find
all sorts of people with other agendas piling on to get the time
interfaces "fixed" for their particular concerns.  Never mind that the
domain of time_t has no logical connection to the leap-second botch;
there are people who will insist that the latter cannot be fixed
unless the committee simultaneously "fixes" the Y2038 problem.  Other
people will want to get reentrant timezone handling.  Thankfully,
Bernstein is not on the committee (that I've seen) so at least there's
nobody insisting that the only possible correct approach is to import
libtai.

-GAWollman