Short of a full time and date library, a single procedure
returning the epoch time in seconds, possibly as a real number, would be useful and is sufficient to implement a full library (though access to the host system's timezone would be desirable in that case).
WG1 voted to provide an optional current-seconds procedure.
Which epoch was left unspecified in the winning ticket, so should be unspecified in the standard.
Superseded by #137.
Note for systems with no clock or concept of time, this could be made optional. Alternately, it could simply be a counter, sufficient to show relative ordering of events but not exact durations.