Make a way to use local device's leap second information #61

Activity

1. rmathar commented on Jun 17, 2026

   

   rmathar

   on Jun 17, 2026

   Last edited by rmathar

   More actions

   There are two scenarios:
   (i) the computer is NOT maintained (because the standard observers fear that
   something might be broken if they upgrade). Typical example are CAHA
   instrument computers that are never updated once the instrument has
   passed the commissioning stage (and where people, even seasoned
   astronomers, are unaware of leap-seconds). In that case it would be vain
   to expect anyone to update anything within the erfa domain.
   (ii) the computer is maintained. In that case further leap seconds trigger
   an update of the date.c source code in the SOFA release, which sooner
   or later triggers a new erfa release with the additional leap seconds,
   and the computer's administrator upgrades to the new erfa. In that case
   my code in #43 would have the
   minimal advantage of perhaps becoming aware of leap seconds in the operating
   system slightly (few months) earlier, but one would need to find a place
   within the erfa code where the function eraDatTZ of #43
   is actually called so the functions benefit from the (potentially early) update of
   the leap-seconds.txt file of the operating system.

   I doubt we could convince the IAU people to implement the
   lookup of the leap seconds in the operating system in their code. They are aware
   that new SOFA releases will occasionally be compiled on very old
   operating systems with nasty side effects to their reputation
   if it turns out that the SOFA software emits wrong results based
   on obsolete leap seconds tables. It is also obvious that my function
   eraDatTZ is based on standards that may not be valid on Windows or MAC's, so if the
   function does not find a leap-seconds.txt, one would still need to have the frozen
   data base of leap seconds in the source code as a backup.

   Useful information is in https://unix.stackexchange.com/questions/806437/
2. mhvk commented on Aug 23, 2026

   

   mhvk

   on Aug 23, 2026

   Contributor

   More actions

   Coming back to this after a long time, I now think that a library reaching into a computer's files is not quite right; liberfa should not do any IO, but just provides an interface to update the leap-second tables that one or more layers up can decide to use. I think that means astropy or perhaps a python helper in pyerfa, i.e., a place where fallbacks and warnings are much easier to organize.