Re: [PoC] Let libpq reject unexpected authentication requests - Mailing list pgsql-hackers

From Jacob Champion
Subject Re: [PoC] Let libpq reject unexpected authentication requests
Date
Msg-id CAAWbhmi4V9zEAvfUSCDFx1pOr3ZWrV9fuxkv_2maRqvyc-m9PQ@mail.gmail.com
Whole thread Raw
In response to Re: [PoC] Let libpq reject unexpected authentication requests  (Jacob Champion <jchampion@timescale.com>)
Responses Re: [PoC] Let libpq reject unexpected authentication requests
List pgsql-hackers
On Fri, Jun 24, 2022 at 12:17 PM Jacob Champion <jchampion@timescale.com> wrote:
> Both NOT (via ! negation) and "none" are implemented in v4.

v5 adds a second patch which implements a client-certificate analogue
to gssencmode; I've named it sslcertmode. This takes the place of the
require_auth=[!]cert setting implemented previously.

As I mentioned upthread, I think sslcertmode=require is the weakest
feature here, since the server always sends a certificate request if
you are using TLS. It would potentially be more useful if we start
expanding TLS setups and middlebox options, but I still only see it as
a troubleshooting feature for administrators. By contrast,
sslcertmode=disable lets you turn off the use of the certificate, no
matter what libpq is able to find in your environment or home
directory. That seems more immediately useful.

With this addition, I'm wondering if GSS encrypted transport should be
removed from the definition/scope of require_auth=gss. We already have
gssencmode to control that, and it would remove an ugly special case
from the patch.

I'll add this patchset to the commitfest.

--Jacob

Attachment

pgsql-hackers by date:

Previous
From: "Ma, Marcus"
Date:
Subject: Retrieving unused tuple attributes in ExecScan
Next
From: Andres Freund
Date:
Subject: Re: Retrieving unused tuple attributes in ExecScan