"Jim C. Nasby" <jnasby@pervasive.com> writes:
> Basically, numbers talk. If there were convincing numbers for something
> that wasn't a corner-case that showed a marked improvement then there'd
> be much more interest in getting this into the backend in some fashion.
I've got no doubt that there are *some* "non corner" cases for which
this would be a win. The lower the update load on the database, the
better it's going to look. The issue really is where does it start
being a loss, and can you convince us that those cases are all "corner"
ones?
(The subtext here, of course, is the assumption that it's an
all-or-nothing choice. I'm of the opinion that supporting both options
would be infeasible from a code complexity and maintenance standpoint;
but a simple patch that did both would of course prove that opinion
wrong ...)
regards, tom lane