Re: constraints and sql92 information_schema compliance

Josh Berkus <josh(at)agliodbs(dot)com> writes:
> 2) Modify the newsysviews to be extensions of the information_schema views:
> e.g. information_schema.tables would have the SQL03 information, and
> information_schema.tables_pg would have pg-specific stuff like table size
> and last analyzed date.

No way. The entire point of information_schema is that it is standard;
adding non-spec things to it renders it no better than direct access
to the PG catalogs.

This thread is fairly interesting since we appear to be watching the SQL
committee allowing a brain-dead choice in the initial information_schema
design to force a non-backwards-compatible dumbing-down of the main spec.
Which they would surely never have done if it weren't for their self-
imposed rules about never changing information_schema (rules that they
appear to follow only erratically anyway ;-))

I'm disinclined to risk being put in a similar bind ... so even if
we were at liberty to put PG-specific stuff into information_schema,
I wouldn't do it.

regards, tom lane