"Pavel Stehule" <[EMAIL PROTECTED]> writes:
> Nobody worked on this subject a lot of years. Who need it, knows about it 
> and use it. I am for this topic is documented and in doc is note about 
> difference from standard and about some incompletness of this operator.

Once we document it there's going to be a backward-compatibility issue
for any proposals to fix it.  And I do think it is broken in a number
of ways:

Per the link Joe cited, I'm not real thrilled with the spec's special
treatment of NULL, but I'd like to see us fix IS OF so that it can be
a substitute for testing tableoid to see which child table a tuple
came from.  And we definitely need to consider the domain issue.

                        regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?


Reply via email to