Things to do:

   * Improve unit test coverage.
     qv: http://cpancover.com/latest/Time-TAI-Simple-1.11/index.html

   * Fix 000-time-tai-simple.t so it doesn't require hand-editing every time there's a new leapsecond.

   * The dependency on POSIX::RT::Clock is preventing Windows testers from testing this module.
     Figure out the Win32-land alternative and use that.
     How to convey to CPAN testers that some dependencies are for specific platforms?
     qv: irclog.freenode-cpantesters 2017-02-17 14:34:58
     qv: https://metacpan.org/source/LEONT/POSIX-RT-Timer-0.017/Build.PL#L9-12

   * Instead of [Prereqs] in dist.ini, use [Prereqs::FromCPANfile] and don't list prereqs there.
     This means I can just maintain the prereqs in cpanfile and eliminate some redundancy.
     qv: .irclog.freenode-perl 2016-03-22 19:28:39

   * I've subscribed to the IERS Bulletin C for advance notification of upcoming leapseconds,
     but it's getting lost in the other mail traffic.  Need to rig something that gets my attention,
     so I know to update the leapsecond list that ships with the module.

   * Add some smarts to occasionally check for clock skew and introduce creeping adjustments.
     Some clocks have a lot of skew.  Most generic motherboard crystals are rated to about
     50ppm, so may drift up to 4 seconds per day, and even very good expensive crystals are only
     rated to 2ppm.  Skew rate may be influenced by transient temperature and ground plane state,
     so need to be careful about assumptions.
     qv: .icbmlog.ttk 2016-03-22 19:43:39