Tom Lane <tgl@sss.pgh.pa.us> writes:
> I think they may have intended to treat each time interval
> as the half-open interval [S,T), that is S <= time < T. However
> that would leave a zero-length interval as completely empty and
> thereby arguably not overlapping anything ... which they didn't
> make it do.
Well an empty interval would be just as inconsistent. I guess they wanted
half-open intervals and they had a problem with zero-length intervals one way
or the other and just chose the one that seemed most useful.
--
greg