Andreas,
> There are only two choices: Creating a minimal subset tool, which will
> rely on INFORMATION_SCHEMA (or a schema API as in ODBC) as standardized
> by SQL specs, or making it specifically for every DBMS, whether using
> some fancy views or not.
Thing is, INFORMATION_SCHEMA doesn't hold a lot of information that people
need to know. Like permissions, comments, object owners, functions, types,
etc. If adding columns and views to the Information schema ... and changing
keys in a couple of places ... is OK, then we have somewhere to go.
Unfortunately, PostgreSQL does not have a seat on the ANSI committee, so we're
not going to get the standard changed. The standard lately belongs to
Oracle and DB2 and we have to suffer under it.
> Doing it seriously, it probably needs the internal DBMS object
> identifiers (oid in the case of pgsql), to uniquely identify objects
> even after a rename. Hiding the OIDs in schema views will reduce their
> usability.
Hmmm ... we argued about this. I was in favor of hiding the OIDs because OIDs
are not consistent after a database reload and names are. I can see your
point though; what do other people think?
--
--Josh
Josh Berkus
Aglio Database Solutions
San Francisco