Re: Vote totals for SET in aborted transaction

Thomas Lockhart <lockhart@fourpalms.org> writes:
> Rather than dismissing this out of hand, try to look at what it *does*
> enable. It allows developers to tune specific queries without having to
> restore values afterwards. Values or settings which may change from
> version to version, so end up embedding time bombs into applications.

I think it's a great idea.  I just want it to be a different syntax from
the existing SET, so as not to break existing applications that expect
SET to be persistent.  It seems to me that marking such a command with
a new syntax is reasonable from a user-friendliness point of view too:
if you write "LOCAL SET foo" or some similar syntax, it is obvious to
every onlooker what your intentions are.  If we redefine "SET" to have
context-dependent semantics, I think we are just creating a recipe for
confusion.
        regards, tom lane