Jeremy Finzel <finz...@gmail.com> writes:
> We often prefer to use timestamptz or "timestamp with time zone" in our
> environment because of its actually storing "objective time" with respect
> to UTC. But in my own work experience, I have scarcely encountered a case
> where business users, and software engineers, do not actually think it
> means the opposite.
Yeah, it's confusing :-(.
> I do believe this is part of the SQL standard,
Actually not, or at least the standard thinks the type should behave
differently from this. They think it should store a local timestamp and
then separately a GMT offset. The way it works in PG was dreamed up ~20
years ago by Tom Lockhart, and in retrospect it was not a great idea
for him to have used the SQL-standard type name for something with
not-SQL-standard behavior.
However, at this point, it's not clear how we could change it without
causing truly enormous backwards-compatibility pain. If it were a
fringe-usage type, maybe we could get away with a renaming, but it's
certainly not that.
> So it seems to me that "timestamp with time zone" is a misnomer in a big
> way, and perhaps it's worth at least clarifying the docs about this,
Don't the docs describe the behavior pretty clearly already?
regards, tom lane
