I've reviewed the PR a few weeks ago. +1 on the new privileges. Thanks for
raising it here, Adnan!

Yufei


On Mon, Nov 16, 2026 at 8:19 PM Adnan Hemani via dev <[email protected]>
wrote:

> Hi all,
>
> Following up on our previous thread about OpenLineage Integration in
> Polaris (
> https://lists.apache.org/thread.html/j4twyvys3g84p1t3lyx0jj8xzwj2b7kv),
> this is a new thread to specifically discuss the new privileges introduced
> as part of PR #5574 <https://github.com/apache/polaris/pull/5574>.
>
> This introduces 4 new privileges: TABLE_READ_LINEAGE,
> TABLE_INGEST_LINEAGE,  VIEW_READ_LINEAGE,  and VIEW_INGEST_LINEAGE - which
> are added to the existing privilege enums. The naming and subtype scoping
> follow the conventions used for the Metrics API's privilege model. The
> following are the semantics related to these privileges:
> * The "READ" lineage privileges allow users to read the lineage information
> for an entity, given that they also have the corresponding "READ"
> privileges on the entities that are being enumerated as a source or
> destination of the requested entity.
> * The "INGEST" lineage privileges allow the user to submit lineage data for
> this entity. When a lineage event cites a table or view as an input, the
> caller needs only read access to it (TABLE_READ_PROPERTIES /
> VIEW_READ_PROPERTIES), not a lineage privilege. Otherwise, a job that reads
> one table and writes another could not record the edge, because ingest is
> implied only by write grants.
>
> Read lineage privileges do not imply ingest lineage privileges, nor do
> ingest lineage privileges imply read lineage privileges. PR #5574 is only
> for adding the privilege model - enforcement will come in a later PR.
>
> Please feel free to chime in your thoughts on this new extension of the
> privilege model!
>
> Best,
> Adnan Hemani
>

Reply via email to