> >I'm curious how difficult this would be to implement...
> The current implementation has the benefit that when you add a column
> existing tuples don't have to be touched.
Hmm, I was under the (obvious) (mis)understanding that a tuple was a
record. Is that not the case? If not, what is it, then?
> Each tuple header stores the number of attributes (natts) at the time of
> its creation. If you query for an attribute with a higher number, you get
> NULL. I don't think this can be changed easily without breaking lots of
> things.
How do the new columns fit into the above scheme?
Chris