On Sat, 2004-12-18 at 13:16, D'Arcy J.M. Cain wrote:
> On Sat, 18 Dec 2004 14:01:04 +0100
> Marc Recht <marc@geht.de> wrote:
> > > I would like to commit the fix I suggested in this PR. I don't know
> > > if it is the perfect fix (see my comments) but if fixes this PR and
> > > also fixes the py-postgresql bulk build.  If no one has a problem I
> > > will commit the changes before the freeze is over.
> > >
> > >
> > 
> > FWIW, I'd like to get rid of the postgresql*-client packages all 
> > together and integrate them into the according libs package.
> 
> Yah, I never really understood splitting them.  After all, if you are
> going to build the libs, you probably want to use them somehow which
> means either also building the server or the client.  If you build the
> client it depends on the server so that's everything.

Hmm? The client depends on the server? There are plenty of occasions
where I want the client and don't need/want the server.

>   If you build the
> server, chances are you want the client in order to admin it and that's
> everything again.  I am sure that there is some theoretical situation
> that requires pieces but I can't imagine that it would be common
> practice.

A webserver accessing an external database by either postgresql libs
directly or through Perl/DBI or PHP, etc. only needs the libs to work.
Fairly common practice I would've thought.

I guess having the client binaries in there is no biggy but I appreciate
the split packages. Certainly their Linux RPMs are split this way too
from a POV of least surprise.

Matt