To: Stef <svb@ucs.co.za>
Cc: Rod Taylor <rbt@rbt.ca>, pgsql-admin@postgresql.org,
	pgsql-sql@postgresql.org
Subject: Re: [SQL] Table versions 
In-reply-to: <1067435458.99917.34.camel@jester> 
References: <20031029125752.35069e3b.svb@ucs.co.za>
	<1067435458.99917.34.camel@jester>
Comments: In-reply-to Rod Taylor <rbt@rbt.ca>
	message dated "Wed, 29 Oct 2003 08:50:58 -0500"
Date: Wed, 29 Oct 2003 09:30:50 -0500
Message-ID: <3731.1067437850@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Rod Taylor <rbt@rbt.ca> writes:
>> What I did next, is put a trigger on pg_attribute that should, in theory,
>> on insert and update, fire up a function that will increment a version

> System tables do not use the same process for row insertion / updates as
> the rest of the system. You're trigger will rarely be fired.

s/rarely/never/.  We do not support triggers on system catalogs.  The
system should have done its best to prevent you from creating one ...
I suppose you had to hack around with a "postgres -O" standalone backend?

Returning to the original problem, it seems to me that comparing "pg_dump
-s" output is a reasonable way to proceed.  The problem of inconsistent
output format across pg_dump versions is a red herring --- just use a
single pg_dump version (the one for your newest server) for all the
dumps.  Recent pg_dump versions still talk to older servers, back to 7.0
or thereabouts.

			regards, tom lane