Re: Granting SET and ALTER SYSTE privileges for GUCs

> On Mar 24, 2022, at 12:06 PM, Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
>
>
> On 3/24/22 12:49, Mark Dilger wrote:
>>
>>> On Mar 17, 2022, at 8:41 AM, Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
>>>
>>> If we abandoned that for this form of GRANT/REVOKE I think we could
>>> probably get away with
>>>
>>>
>>> GRANT { SET | ALTER SYSTEM } ON setting_name ...
>>>
>>>
>>> I haven't tried it, so I could be all wrong.
>> Version 12 of the patch uses SET and ALTER SYSTEM as the names of the privileges, and PARAMETER as the name of the thing on which the privilege is granted. The catalog table which tracks these grants is now named pg_parameter_acl, and various other parts of the patch have been adjusted to use a "parameter" based, rather than a "setting" based, naming scheme. One exception to this rule is the "setacl" column in pg_parameter_acl, which is much more compact than the "parameteracl" name would be, so that remains under the old name.
>
>
> I can live with it I guess, but it seems perverse to me to have
> pg_settings but pg_paramater_acl effectively referring to the same set
> of things. If we're going to do this perhaps we should create a
> pg_parameters view which is identical to pg_settings and deprecate
> pg_settings. I don;t want to hold up this patch, I think this can
> probably be managed as a follow up item.

Right, the version 12 patch was following Peter's and Tom's comments upthread:

> On Mar 17, 2022, at 7:47 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> Well, I still say that "SET" is sufficient for the one privilege name
> (unless we really can't make Bison handle that, which I doubt). But
> I'm willing to yield on using "ALTER SYSTEM" for the other.
>
> If we go with s/SETTING/PARAMETER/ as per your other message, then
> that would be adequately consistent with the docs I think. So it'd
> be
>
> GRANT { SET | ALTER SYSTEM } ON PARAMETER foo TO ...
>
> and the new catalog would be pg_parameter_acl, and so on.

We could debate that again, but it seems awfully late in the development cycle. I'd rather just get this committed, barring any objections?

—
Mark Dilger
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company