Brendan Jurd <direvus@gmail.com> writes:
> If the current implementation of timenow() is truly obsolete, would it
> be verboten to change its return type? We could rewrite the function
> to return timestamp, for example.
[ shrug... ] This is just a variant of the choose-a-new-function-name
game. If we are going to choose a new function name, choosing one that
collides with an existing name (obsolete or not) doesn't seem like a
win to me. You could just as well choose another name, and avoid
angering whoever out there might still be using timenow().
BTW: at least with our current interpretation of these datatypes, the
only type that is sensible for a now()-like function to return is
timestamptz. Not plain timestamp; that cannot be considered to
represent a well-defined instant at all.
regards, tom lane