On Wed, Jan 10, 2024 at 2:27 PM Dagfinn Ilmari Mannsåker
<ilmari@ilmari.org> wrote:
>
> Magnus Hagander <magnus@hagander.net> writes:
>
> > On Wed, Jan 10, 2024 at 1:44 PM Aleksander Alekseev
> > <aleksander@timescale.com> wrote:
> >>
> >> It hurts my sense of beauty that usename and authname are of different
> >> types. But if I'm the only one, maybe we can close our eyes on this.
> >> Also I suspect that placing usename and authname in a close proximity
> >> can be somewhat confusing. Perhaps adding authname as the last column
> >> of the view will solve both nitpicks?
> >
> > But it should probably actually be name, given that's the underlying
> > datatype. I kept changing it around and ended up half way in
> > between...
>
> https://www.postgresql.org/docs/current/functions-info.html#FUNCTIONS-INFO-SESSION-TABLE
> (and pg_typeof(system_user)) says it's text. Which makes sense, since
> it can easily be longer than 63 bytes, e.g. in the case of a TLS client
> certificate DN.
We already truncate all those to NAMEDATALEN in pg_stat_ssl for
example. so I think the truncation part of those should be OK. We'll
truncate "a little bit more" since we also have the 'cert:', but not
significantly so I think.
but yeah, conceptually it should probably be text because name is
supposedly a *postgres identifier*, which this is not.
--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/