On May 11, 2005, at 7:38, Greg Sabino Mullane wrote:
> So they are willing to learn the new system views, but not the system
> tables? The above seems an argument for I_S, or at least an expanded
> I_S.
>
> So... the reason we don't want to expand (not alter) I_S is that it is
> a
> "standard" that very few RDBMS actually bother to implement, is already
> out of date, and is incomplete? Seems we bend the rules in other ways
> when
> needed (e.g. lowercase relation names), we could certainly add
> additional
> tables and columns here, while maintaining the "standard" set for
> applications
> looking for them.
>
One of the reasons I've been impressed with PostgreSQL and its
developers is that I've seen respect for the SQL specifications *except
in cases where it would seriously break backwards compatibility*. In
implementing new features, if the SQL spec has something to say about
it, it's been my observation that good efforts have been made to
comply, though sometimes other syntax or PostgreSQL extensions are
made.
This is not to say the SQL spec is perfect. (At heart I lean toward the
Date/Darwin relational model, but that's just me :) However, to take
something that *is* specified by SQL (and if I understand correctly,
was *implemented in PostgreSQL specifically for SQL compliance*, it
would be a shame to break that. I think PostgreSQL's spec compliance is
a nice bragging point as well -- we do the spec, and more :)
Additional views that depend where possible on the INFORMATION_SCHEMA
could actually be a good thing, as the INFORMATION_SCHEMA follows the
spec, it'd be less likely to change between versions and make
maintenance easier. That said, I haven't looked at the work the new
systems views people have done. I recognize their motivation, as the
times I've needed to look at the current system tables, it's always
been with the docs open right beside me, flipping between pages to see
everything I need to join to get the information I want. I for one am
happy and grateful that a group of people have taken it upon themselves
to provide an easier way to view Postgres system information, and think
that the additional views in some form would make a great addition to
PostgreSQL.
Michael Glaesemann
grzm myrealbox com
