On Tuesday 27 January 2009 17:19:28 Tom Lane wrote:
> If we'd had the SQL-spec behavior from day one, it wouldn't be a
> problem, but you can't just blow off the old behavior like that.
> It's a potential security hole, since GRANT ALL on a view used to
> be de facto the same as GRANT SELECT, if you hadn't bothered to
> create any rules.
That is a good point. But the only clean solution would be to make views
never updatable by default, and invent a nonstandard syntax to make them so,
which seems very unattractive to me. A GUC variable as a transition measure
could work, though.