Thomas Lockhart <lockhart@alumni.caltech.edu> writes:
> I'd be a little reluctant to give up the alternate "::" syntax, only
> because I'm not sure I trust the standards folks to not stomp on the
> alternative at some point in the future.
Even more to the point, we now have a substantial pool of user
applications that rely on "::" (I know a lot of my company's
code does, for example). We can't just blow off those users.
Peter's current proposal seems OK to me, with the caveat that we
will have to be *very* circumspect about introducing additional
variables-automatically-defined-by-psql in the future. Every
time we do, we risk breaking existing user scripts, not unlike
what happens when we introduce a new reserved word in the backend
grammar.
Lane's Other Law: the price of success is backwards compatibility.
regards, tom lane