Re: Controlling changes in plpgsql variable resolution

Robert Haas <robertmhaas@gmail.com> writes:
> I actually think that we should not have a GUC for this at all.  We
> should have a compiled-in default, and it should be error.  If you
> want some other behavior, decorate your functions with #option.

We've agreed that the factory default should be "error", but I don't
think we can go much further than that in the direction of breaking
peoples' code.  We need to provide a backwards-compatible option,
IMHO, so that this isn't seen as a roadblock to updating to 8.5.
(You can argue all you want about whether it really is one, but it'll
be seen as one.)  And the Oracle-compatible option will be attractive
to people coming in from that side.  Reviewing megabytes of pl/sql
code for this kind of gotcha is not fun, and the "error" default would
only help a bit.

One advantage of making the GUC be SUSET is that those who want to take
a paranoid approach here have an easy answer: don't allow it to be
changed from "error".  We are allowed to assume that those who do change
it are responsible adults.
        regards, tom lane