Re: pg_proc?

2014-12-17 8:55 GMT+01:00 David G Johnston <david(dot)g(dot)johnston(at)gmail(dot)com>:
>
> Guillaume Lelarge-3 wrote
> > Hi,
> >
> > Le 8 déc. 2014 20:53, "Wim Bertels" &lt;
>
> > wim.bertels@
>
> > &gt; a écrit :
> >>
> >> Hallo,
> >>
> >> pg_admin needs access to pg_proc of a database to work,
> >> it fails without having this access, (wanting to read out the metadata)
> >> while eg psql succeeds
> >>
> >> use case:
> >>
> >> revoke all on pg_catalog.pg_proc from public;
> >>
> >> then try to connect with pgadmin.
> >>
> >> for non technical users this poses a hurdle,
> >> hence the feature suggestion: don't show the metadata, just open a
> >> pgadmin sql window in that case
> >>
> >
> > Can you be more explicit? When do you want that PgAdmin opens a SQL
> > window,
> > and why?
>
> Without investigating I think the request can be summed up as:
>
> Treat the failure to read meta-data as a warning, not an error. Continue
> on
> doing whatever main activity the user is engaged in and pretend that
> whatever meta-data queries were run simply returned zero records instead of
> an error.
>
> I can support that general sentiment though in this specific case I'm not
> sure what other downsides there are to revoking pg_proc access to conclude
> that it is a valid use-case to be catered to.
>
> Since psql doesn't preemptively load meta-data it doesn't have the same
> problem - though using the appropriate meta-queries would (I think) result
> in the same error being provoked there as well.
>
>
Yeah, the only differences between psql and pgAdmin is that, most of the
time, pgAdmin asks for the metadata itself, whereas you need to ask psql to
give it to you.

On treating failure as a warning, not an error, it'd be quite hard to do it
right, if it's at all possible. But I'd quite happy to see a patch adding
this capacity.

--
Guillaume.
http://blog.guillaume.lelarge.info
http://www.dalibo.com