From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>
cc: pgsql-hackers@lists.postgresql.org
Subject: Re: Fuzzy thinking in is_publishable_class
In-reply-to: <15150.1557257111@sss.pgh.pa.us>
References: <15150.1557257111@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Tue, 07 May 2019 15:25:11 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2660.1557264605.1@sss.pgh.pa.us>
Date: Tue, 07 May 2019 17:30:05 -0400
Message-ID: <2661.1557264605@sss.pgh.pa.us>
Precedence: bulk

I wrote:
> is_publishable_class has a test "relid >= FirstNormalObjectId",
> which I think we should drop, for two reasons:

> 1. It makes the comment claiming that this function tests the same
> things as check_publication_add_relation a lie.

> 2. The comment about it claims that the purpose is to reject
> information_schema relations, but if that's so, it's ineffective.
> We consider it supported to drop and recreate information_schema,
> and have indeed recommended doing so for some minor-version
> upgrades.  After that, the information_schema relations would no
> longer have OIDs recognizable to this test.

> So what is the motivation for this test?  If there's an important
> reason for it, we need to find a less fragile way to express it.

After further digging around, I wonder whether this test wasn't
somehow related to the issue described in

https://postgr.es/m/2321.1557263978@sss.pgh.pa.us

That doesn't completely make sense, since the restriction on
relkind should render it moot whether IsCatalogClass thinks
that a toast table is a catalog table, but maybe there's a link?

			regards, tom lane