Tom Lane wrote:
> Well, that probably knocks out my thought that we could stop reserving
> the system column names (at least ctid and xmin, which are the two that
> actually seem useful to ordinary clients, need to stay reserved). But
> it still seems like we don't have to represent these columns explicitly
> in pg_attribute.
Hm, technically you might be right. Still, I like pgAdmin3 to show that
columns (when "show system objects" is enabled) for teaching purposes,
so users/newbies browsing the objects will learn "hey, there are some
reserved columns, they could have some meaning". I'd be not too excited
about emulating system column pg_attribute entries...
Regards,
Andreas
