Re: User functions for building SCRAM secrets

From: Michael Paquier <michael(at)paquier(dot)xyz>
To: Daniel Gustafsson <daniel(at)yesql(dot)se>
Cc: "Jonathan S(dot) Katz" <jkatz(at)postgresql(dot)org>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: User functions for building SCRAM secrets
Date: 2022-11-30 01:12:00
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

On Tue, Nov 29, 2022 at 09:32:34PM +0100, Daniel Gustafsson wrote:
> On the whole I tend to agree with Jacob upthread, while this does provide
> consistency it doesn't seem to move the needle for best practices. Allowing
> scram_build_secret_sha256('password', 'a', 1); with the password potentially
> going in cleartext over the wire and into the logs doesn't seem like a great
> tradeoff for the (IMHO) niche usecases it would satisfy.

Should we try to make \password and libpq more flexible instead? Two
things got discussed in this area since v10:
- The length of the random salt.
- The iteration number.

Or we could bump up the defaults, and come back to that in a few
years, again.. ;p

In response to


Browse pgsql-hackers by date

  From Date Subject
Next Message Melanie Plageman 2022-11-30 01:12:47 Re: pg_stat_bgwriter.buffers_backend is pretty meaningless (and more?)
Previous Message Michael Paquier 2022-11-30 01:00:55 Re: Collation version tracking for macOS