Experimental Physics and
Industrial Control System

Philip Taylor <[email protected]> · Fri, 24 Mar 1995 10:26:52 +0000 (GMT)

Regarding leap seconds in EPICS timestamps :

The number that is held in the EPICS timestamp (seconds/nanoseconds 
since epoch Jan 1 1990) does *not* actually represent the precise 
elapsed time since the epoch : leap seconds are ignored and so the 
number of seconds in the timestamp is slightly less than the actual 
elapsed time.

It only matters if an elapsed time is calculated from a pair of EPICS 
timestamps and a leap second ocurred during the period between the 
two timestamps. There is also a problem during the leap second 
itself - an ambiguity in the timestamp value which is the same for 2 
successive seconds.

There is nothing unusual about this : this behaviour is identical to 
using a standard Unix (POSIX/ANSI C) time value = "Seconds since Jan 1 
1970". Again, this does not actually represent the precise number of 
seconds that have elapsed since the epoch since leaps are ignored but is 
simply a way of representing time as a single integer, the standard 
routines convert to/from UTC using a simple formula ignoring leaps. 

Apparently there are proposals under way to include leap second handling 
in a defined way in future C and Unix standards but at the moment the 
problem is ignored.

Philip

------------------------------------------------------------------------------------
Philip Taylor, Royal Greenwich Observatory, Madingley Road, CAMBRIDGE, UK.  CB3 0EZ.
Email : [email protected]                              Phone : (UK+) (0)1223-374710
------------------------------------------------------------------------------------

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System