I am currently reading up on the history behind inheritance in postgres
and found this interesting post back in May 2000.
http://archives.postgresql.org/pgsql-hackers/2000-05/msg01349.php
>From reading other posts, it seems that there is the idea that since
under is single inheritance then tables "under" another are all really
in that same relation only they act the as expected to the user. It
seems that this would allow backwards compatibility with current
postgres databases that use inheritance and offers a simplified single
relation inheritance that would fix problems with primary keys
(references don't see id's created via a child table,etc), constraints,
triggers, etc.
http://archives.postgresql.org/pgsql-hackers/2002-08/msg01103.php
http://archives.postgresql.org/pgsql-hackers/2002-09/msg01207.php
Since the old inheritance stuff would be the same as now (does not
inherit primary key, etc.) it would address other concerns as well:
http://archives.postgresql.org/pgsql-hackers/2002-08/msg01099.php
>From a point of view, inheritance is not broken it just works a
different way; it does not violate constraints because the "only" clause
shows that the value is not really in that table. With "under", the
only clause, if it exists at all, could just basically filter for
entries not added to columns from any "tables" added via under and show
only the value directly inserted into the parent columns. Tables under
another are physically one relation, tables inheriting from one another
are physically different relations.
Besides, "under" would be a new feature!
Babak.