🚀 go-pugleaf

alpha@waggen.twuug.com (Joe Wright) writes:
>Please remember that floats (and doubles) are approximates.
>The value of time_t must be precise.  We don't like a value
>which says 'it's almost 5 o'clock'.  Precision requires
>time_t to be an integer.

But time is a continuum; it is not discrete. Integers, floats and doubles all
store discrete values, so none of them can precisely represent all values of
time, not even in a small fixed interval. You may not like a value that says
"it's about 5 oclock", but that's the best we can do. A floating point time_t
will allow finer resolution for times near the epoch than an integral one, and
less resolution for times far from the epoch. An integral time_t abruptly
becomes unable to represent the time at all as you move away from the epoch,
whereas the precision of the floating point time_t degrades gracefully. Since
we are normally more interested in times near the present than the distant past
or future, this is a reasonable tradeoff. I think this was the original
poster's point.

regards,
Graham