Re: [Pgsqlrpms-hackers] Safer auto-initdb for RPM init

On Saturday 26 August 2006 22:08, Matthew T. O'Connor wrote:
> Joshua D. Drake wrote:
> > Matthew T. O'Connor wrote:
> >> script. If we installed the datadir during the RPM install, it would
> >> still be newbie friendly and would removed initdb from start script
> >> solving that problem.
> > initdb will not overwrite an existing installation.

> Poorly chosen words. I meant, the problem where the start script will
> create a new data dir when it doesn't see that one exists even though
> one actually does exist it's just not available at the moment. Either
> way, if the start scripts never created a data dir, then there is no
> problem.

If a prebuilt datadir tree were available, it would make it even easier for
people to overwrite their existing data, as RPM does not check
non-RPM-managed files prior to overwriting them.

This was in fact done several years ago by Red Hat, and was speedily removed.

I understand the needs from both angles; so I'll ask just a simple question:
which is worse, annoying all the newbies who just want to get started
quickly, or annoying the small number of people who NFS mount their datadirs?
(I personally wouldn't in a million years trust NFS for ACID compliance;
maybe iSCSI, but NFS?!?!).

The behavior, in my opinion, should be configurable and ON by default.
--
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC 28772
(828)862-5554
www.pari.edu