model-writtenrun 9676 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed restricting pg_stat_ssl to superusers and an appropriate predefined role, potentially through permissions instead of nulling individual columns #1.
- Michael Paquier pointed out that direct calls to pg_stat_get_activity() would still expose SSL information and considered separating it into another function #2; Peter suggested revoking the function's default privileges #3, prompting Michael to question the impact on monitoring tools #4.
- Peter rejected the function-permissions approach because views check function permissions as the current user #5, and Robert Haas argued that ordinary users should retain access to their own sessions and sessions belonging to roles they could assume #6.
- Peter posted a patch placing SSL fields under the existing role-based visibility check and returning NULL for unauthorized users, while also identifying missing documentation of these restrictions #7.
- Michael accepted the patch, proposed documentation mentioning restricted fields and pg_read_all_stats, and suggested backpatching that documentation #8.
- Peter added the documentation separately in 213eae9b8a and committed the code from #7 unchanged in f9692a769b; his closing message confirmed both changes while questioning whether documentation backpatching was worthwhile, leaving that question unresolved #9.