Tom, all,
* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> It may be that the functions Joel proposes are worth having for other
> tools to use, but I'm not in favor of making pg_dump use them.
I care very little about migrating the pg_dump functionality into
server-side functions and a great deal about *having* such functionality
available server-side. We've been around this time and time again:
there are use-cases beyond pg_dump for being able to get the definition
of an object. Having to first figure out and then replicate what pg_dump
(or psql) does is no trivial feat.
In short, I think I agree w/ Tom here (though I contend that there *are*
such use-cases, not that 'it may be'.. :). Let's drop the discussion
about changing pg_dump and/or psql and instead simply go implement these
functions. We can revisit the pg_dump discussion 5 or 10 years down the
road, after we've seen how well these functions work and what uses they
are put to outside of our (relatively small) world.
Thanks,
Stephen