Also, just how strictly do we have to follow the standard? Requiring sentinels like '01 JAN 3000` just seems so silly. Could Postgres permit nullable start/end PERIOD columns, and give them the same meaning as ranges (unbounded)? Even if I forgot about ranges altogether, I'd sure love to avoid these sentinels.
We have "infinity" and "-infinity" values in our date and timestamp types:
I think this avoids the silliness with sentinel values.
For myself, I don't care about PERIOD etc. one bit. The "every new capability gets its own syntax" model that SQL follows is very old-fashioned, and for good reason. I'm happy with ranges and exclusion constraints. But if we can provide an implementation of PERIOD that makes it easier to port applications written for legacy database systems, it might be worthwhile.
