+1! I still think we should store the whole TableMetadata or carve out a
path forward for it, but this feels directionally correct to me.

On Mon, Feb 2, 2026 at 4:31 PM Yufei Gu <[email protected]> wrote:

> +1 from me, this makes sense.
>
> Avoiding a full loadTable for cases like credential vending is a clear win,
> and keeping the relevant location properties in the table entity feels like
> a reasonable incremental step. I believe it also aligns well with the
> earlier direction of storing selected metadata in persistence without
> committing to persisting the full TableMetadata yet.
>
> Thanks for driving this and for the PR.
>
> Yufei
>
>
> On Tue, Jan 27, 2026 at 9:55 AM Prashant Singh via dev <
> [email protected]> wrote:
>
> > Hello everyone,
> >
> > Presently since not all the location properties are stored in the entity
> > one has to do complete loadTable (i.e read the table from the object
> > store), even for existing use cases such as credential endpoint i.e
> > vend the cred of the table. This is super expensive, we can optimize this
> > if we start keeping these properties of the table in the table entity
> > (polaris persistence).
> >
> > With this we will have a way to do cred vending, in future (to remote
> sign)
> > without going to object store.
> >
> > Note if we take dependency on this we would have to think of *backfill*
> but
> > for step 1 it seems a reasonable step as i personally see this as an
> > extension of storing yet another set of props like we did [1], while we
> > still think / debate on if we wanna store the whole table metadata in
> > persistence, would love to know what other folks think, if anyone have
> > *objection* to it.
> >
> > I proactively have created a pr [2] for the same.
> >
> > Huge thanks to Alex for proposing this in their remote signing effort in
> > the first place.
> >
> > [1] https://github.com/apache/polaris/pull/2735
> >
> > [2] https://github.com/apache/polaris/pull/3226
> >
> > [3] https://github.com/apache/polaris/pull/2280#discussion_r2519568503
> >
> >
> > Best,
> >
> > Prashant Singh
> >
>

Reply via email to