On Thu, 2003-04-24 at 18:28, Tom Lane wrote:
> Sailesh Krishnamurthy <sailesh@cs.berkeley.edu> writes:
> > Rod> The INFORMATION_SCHEMA? Out of curiousity, how do they
> > Rod> handle DB2 extensions? Do they create new views in that
> > Rod> schema? Do they ignore them?
>
> > Why extensions, even for things like indexes that aren't in the
> > standard, they create views (SYSCAT.INDEXES, SYSCAT.INDEXAUTH etc.)
> > ...
> > Certainly - it's just that the meaning and number of existing columns
> > and rows in the syscat views are always backward compatible. That
> > includes support of the info schema - for the sql standard features
> > that db2 supports.
>
> > So if there's something new in the catalog tables that is a result of
> > an extension and doesn't appear as a column in the syscat views (or
> > the info schema) then an appropriate column may be added to the view -
> > provided that this doesn't break the info schema compatibility.
>
> Of course, IBM can afford to keep reps on the SQL standards committee to
> make sure that no future spec extension conflicts with the names they've
> used for their additions to INFORMATION_SCHEMA. We, on the other hand,
> could easily get burnt by spec changes.
We could probably get away with adding pg_ views to the information
schema though. For extensions of an existing view, simply inherit the
real view into a pg_ labelled view and add the new columns.
--
Rod Taylor <rbt@rbt.ca>
PGP Key: http://www.rbt.ca/rbtpub.asc
