> As an implementor I'm always wary of using features nobody else has,
> especially in databases. So, if I'd want postgres to have one thing
> nobody else has, it would be the most complete standard SQL
> implementation - so it would at least be the other products' fault if
> I'd have to do any special porting work to/from postgres.
Why can't both be done? If nobody extended the spec or came up with new
features there wouldn't exactly be any progress.
Yes, meeting the spec is a good goal, and one that is getting quite
close as far as the SQL part goes -- but it shouldn't be the only goal.
Inheritance currently saves me from issuing ~4 inserts, updates, deletes
as it handles it itself. If indexes and a couple other things worked
across the entire tree it could be more useful.
I think what we need to do is expand on it, not blow it away.
There is a list of spec features we support. Stick to those (or the
subset) that is appropriate for portability. If you plan on making an
embedded DB Based application the extra features may be useful.