Date: Sun, 11 Sep 2022 21:46:47 -0700
From: Noah Misch <noah@leadboat.com>
To: Bruce Momjian <bruce@momjian.us>,
	Peter Eisentraut <peter.eisentraut@enterprisedb.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Erki Eessaar <erki.eessaar@taltech.ee>,
	"pgsql-docs@lists.postgresql.org" <pgsql-docs@lists.postgresql.org>
Subject: Re: SQL-standard function bodies and creating SECURITY DEFINER
 routines securely
Message-ID: <20220912044647.GA1404410@rfd.leadboat.com>
References: 
 <AM9PR01MB8268BF5E74E119828251FD34FE409@AM9PR01MB8268.eurprd01.prod.exchangelabs.com>
 <Yvvw1JB/Y12HvAYn@momjian.us>
 <1408823.1660678462@sss.pgh.pa.us>
 <YvvyJTmSbPLKUm5k@momjian.us>
 <YxAG2DFuSVDcQMn9@momjian.us>
 <a2dc9de4-24fc-3222-87d3-0def8057d7d8@enterprisedb.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <a2dc9de4-24fc-3222-87d3-0def8057d7d8@enterprisedb.com>
User-Agent: Mutt/1.5.24 (2015-08-30)
Archived-At: 
 <https://www.postgresql.org/message-id/20220912044647.GA1404410%40rfd.leadboat.com>
Precedence: bulk

On Thu, Sep 08, 2022 at 01:20:31PM +0200, Peter Eisentraut wrote:
> On 01.09.22 03:11, Bruce Momjian wrote:
> >On Tue, Aug 16, 2022 at 03:38:13PM -0400, Bruce Momjian wrote:
> >>On Tue, Aug 16, 2022 at 03:34:22PM -0400, Tom Lane wrote:
> >>>Bruce Momjian <bruce@momjian.us> writes:
> >>>>I have written the attached patch to mention this issue about sql_body
> >>>>functions.
> >>>
> >>>Spell-check, please.  Seems OK otherwise.

> >Patch applied back to PG 10.  Thanks.
> 
> This feature is new in PG 14, so backpatching further than that doesn't make
> sense.

Even an sql_body function should override search_path, because it may call
other code that reacts to search_path.  Separately, the new sentence is near
the start of a section that addresses more than just search_path.  The section
ends with the "revoke the default PUBLIC privileges" topic, which is no less
relevant to sql_body functions.

Documentation needn't explain cases that make a best practice optional, and it
should explain only valuable ones.  Omitting search_path on sql_body SECURITY
DEFINER functions isn't that valuable.  If it were valuable, the patch's
sentence gives too little detail for the reader to decide what's safe for a
given function.  I think this section should not attempt such detail.  It's
enough to give the best practice, as the documentation did before this change.