Re: exposing pg_controldata and pg_config as functions

On Mon, Aug 31, 2015 at 9:47 PM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
> On 8/25/15 11:32 PM, Joe Conway wrote:
>> 1.) pg_controldata() function and pg_controldata view added
>
> I don't think dumping out whatever pg_controldata happens to print as a
> bunch of text fields is very sophisticated. We have functionality to
> compute with WAL positions, for example, and they won't be of much use
> if this is going to be all text.
>
> Also, the GUC settings tracked in pg_control can already be viewed using
> normal mechanisms, so we don't need a second way to see them.
>
> The fact that some of this is stored in pg_control and some is not is
> really an implementation detail. We should be thinking of ways to
> expose specific useful information in useful ways, not just dump out
> everything we can find. Ultimately, I think we would like to move away
> from people parsing textual pg_controldata output, but this proposal is
> not moving very far in that direction.

The nice thing about dumping the information as text is that you can
return every value in the same universal format: text. There's a lot
to like about that.

But I'm not sure I like the idea of adding a server dependency on the
ability to exec pg_controldata. That seems like it could be
unreliable at best, and a security vulnerability at worst.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company