Re: [HACKERS] DROP TABLE inside a transaction block

The Hermit Hacker wrote:
> IMHO, we are basically supporting ppl *not* doing regular backups of their
> data ... most, if not all, of the problems that ppl feel exist that
> requires the use of 'flat files', IMHO, aren't big problems if properly
> backup procedures are followed ...

I suggested the 'flat-file' more as a compromise than anything else.
(Although it kindof backfired :-().  Technically speaking, my
'flat-file' is trading the flat-file in the OS's filesystem (the
directory) with a separate flat-file.  Little to no admin difference
from my point of view.

The problem that Bruce is talking about occurs when you try to restore
(from a properly built off-line binary backup) a single table or small
set of tables.  It doesn't have anything to do with supporting people
who won't do proper backups, IMO.

Of course, I personally use on-line pg_dump backups and feed into psql
for on-line restore -- which doesn't require knowing anything about the
underlying filesystem structures.

So, the dichotomy is between those who want to admin at the OS file
level versus those who feel the backend should hide all those details
from the admin.  At least that's how istm.

--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11