Re: [HACKERS] Re: [SQL] Proposed Changes to PostgreSQL

<I trimmed the CC list a bit>

Chris Bitmead wrote:
>
> Mark Hollomon wrote:
> >
> > > [ discussion on changing the default to getting subclasses ]
> >
> > I object.
>
> Tell me why you object. Performance concerns? Compatibility?

Definitely compatibility. The load I see (200 - 300 queries a DAY)
isn't enough for me to be concerned about an extra millisecond
or two per query. But I certainly understand others concerns in
this area.

One of my responsibilities at work is the maintenance of a homegrown
document indexing and retrieval system. It is about 100K of Perl
that calls into a custom Perl wrapper around libpq. The system
is an escaped 'proof-of-concept'. I wrote it using inheritance
features of Postgres95.

The upshot is, that this proposed change would require me to examine
almost every line of this system in order to make sure that I put
ONLY in just the right spots. Yes, this would be where ever there
_isn't_ a '*', but how do I grep for the lack of a asterisk? Since
it is a "prototype", The code feels very free to pass around small
snippets of SQL, a disembodied FROM clause, a portion of a VALUES
clause.

I simply would not be allowed the time to do the rewrite necessary
to accomodate this change. And if I _did_ have the time, I would
probably rewrite it for Oracle because then DB Admin would be someone
_else's_ job.

Now, one of the days, I will find a good excuse (eg new feature)
to do a complete rewrite. And _then_ your proposal will actually
be a help.

And that is why I suggest a SET variable. When I'm ready to
use the new feature, I can. But no work is necessary until that
day arrives.

Thanks for listening.

--

Mark Hollomon
mhh(at)nortelnetworks(dot)com
ESN 451-9008 (302)454-9008