JB, Dennis, and I have raised several concerns in the
thread(https://github.com/apache/polaris/pull/5246/changes#r4383768595),
and it looks like they are still unresolved. The original comments
date back to Oct 26, but we haven't received any responses or
follow-ups for over a week.

Dmitri, could you please check with the author and help get these
concerns addressed?

Yufei

On Thu, Nov 5, 2026 at 6:15 AM Dmitri Bourlatchkov via dev
<[email protected]> wrote:
>
> Hi Yufei,
>
> I believe the comment thread you mentioned [1] has been resolved for
> relevant grants.
>
> If you think something is missing, please respond with a more specific
> rationale.
>
> PR [5246] has several approvals. As I mentioned before, I think it is
> essential to have it in 1.9.0 to prepare users for future migration toward
> better authorization coverage. Delaying this PR by another release will
> effectively delay authorization improvements on the user side by two (!)
> releases.
>
> Please note that the main purpose of this PR is establishing the new AuthZ
> operation and a feature flag. The new check is _not_ enforced by default.
> Any remaining changes for Ranger and internal RBAC grants can be done on
> main after 1.9.0.
>
> I propose merging it before 1.9.0 is branched, ideally today or tomorrow.
>
> Thanks,
> Dmitri.
>
> [5246] https://github.com/apache/polaris/pull/5246
>
> [1] https://github.com/apache/polaris/pull/5246/changes#r4383768595
>
> On Wed, Nov 4, 2026 at 6:52 PM Yufei Gu <[email protected]> wrote:
>
> > I'd suggest not merging when there are comments needed to be resolved,
> > here is an open comment:
> > https://github.com/apache/polaris/pull/5246/changes#r4383768595
> > Yufei
> >
> > On Wed, Nov 4, 2026 at 3:45 PM Dmitri Bourlatchkov via dev
> > <[email protected]> wrote:
> > >
> > > Hi All,
> > >
> > > Given an extensive review period, I propose merging [5246] on Nov 6.
> > >
> > > Note: this PR does _not_ introduce any behaviour changes. It only
> > prepares
> > > options for stricter authZ in the future.
> > >
> > > Please reply to this thread or comment in GH if you have any concerns.
> > >
> > > Thanks,
> > > Dmitri.
> > >
> > > [5246] https://github.com/apache/polaris/pull/5246
> > >
> > > On Mon, Nov 2, 2026 at 10:16 AM Dmitri Bourlatchkov <[email protected]>
> > > wrote:
> > >
> > > > Hi Vignesh,
> > > >
> > > > Thanks for promptly addressing the feedback!
> > > >
> > > > I think PR [5246] approaches a mergeable state.
> > > >
> > > > Given that it introduces a migration action for users, which is
> > optional
> > > > in the current code, I think it makes sense to include the PR into the
> > > > upcoming 1.9.0 release. This will enanle working on flipping the flag
> > In
> > > > the following release.
> > > >
> > > > Yufei, Dennis: Please take a look at the latest PR state.
> > > >
> > > > [5246] https://github.com/apache/polaris/pull/5246
> > > >
> > > > Thanks,
> > > > Dmitri.
> > > >
> > > >
> > > > On Sat, Oct 31, 2026 at 12:11 PM vignesh a <[email protected]>
> > wrote:
> > > >
> > > >> Hi all,
> > > >>
> > > >> Thanks Yufei, Dmitri and JB. I've restacked #5246 [1]
> > > >>
> > > >> Dropped GET_CATALOG_CONFIG_PROPERTIES. That takes out the second
> > > >> authorizer op and the partial-result case Yufei called out, where a
> > > >> client can't tell absent settings from withheld ones.
> > > >>
> > > >> Kept GET_CATALOG_CONFIG behind ENFORCE_CATALOG_CONFIG_AUTHORIZATION
> > > >> (default false), as Dmitri asked. That covers the /config AuthZ gap as
> > > >> of 1.8.0, and existing deployments keep today's behaviour.
> > > >>
> > > >> Once that check passes, the response still returns catalog properties
> > > >> as it does today.
> > > >>
> > > >> CHANGELOG says the default is not flipping until the client-facing
> > > >> config follow-up is settled, as JB asked. Client-facing vs server-only
> > > >> properties can stay in its own thread.
> > > >>
> > > >> PR is updated now. PTAL
> > > >>
> > > >> Thanks,
> > > >> *Vignesh*
> > > >>
> > > >> [1] https://github.com/apache/polaris/pull/5246
> > > >>
> > > >
> >

Reply via email to