model-writtenrun 9684 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- John Naylor proposed generating TOAST declarations and assigning OIDs automatically for non-shared catalogs, while retaining stable explicit OIDs for shared catalogs, in #1.
- Tom Lane objected that stable catalog OIDs help forensic investigation in #2, Stephen Frost questioned the benefit in #3, and Andres Freund argued that automatic assignment would remove a manual catalog-creation step in #4.
- The discussion shifted to existing genbki allocation when Tom Lane warned that generated OIDs could collide with initdb allocations and consume the 9000-range OIDs used by forks in #5; he preferred separate ranges with 2000 OIDs for genbki over a dynamically adjusted bootstrap boundary in #7.
- John Naylor and Andres Freund disputed whether automatic assignment introduced a new policy problem, pointing out that missing explicit OIDs had already been possible, in #8 and #9.
- Andres Freund supplied a separate patch reserving OIDs 10000-11999 for genbki and starting bootstrap allocation at 12000, updating function lookup tables, postgres_fdw and unused_oids while leaving dependency-related cutoffs unchanged, in #9; he retained FirstGenbkiObjectId after rejecting John Naylor's broader naming suggestion in #10 and #11.
- Andres Freund committed #9 as 09568ec3d3, adding only the Windows build invocation adjustment described in #12.
- After the commit, Tom Lane requested an allocation-boundary check, questioned dependency cutoff assumptions, and proposed treating low OIDs as implicitly pinned to reduce pg_depend entries in #13; John Naylor identified template1 as another pinning exception in #14, Andres Freund discussed checks and pin creation in #15, and Alvaro Herrera suggested moving the public schema OID outside the special range in #16.
- Two months later, John Naylor submitted the comment correction in #18, which Michael Paquier applied unchanged as f0cce9fcb5 and confirmed in #19.
- The thread ends with Andres Freund's acknowledgment in #20: the OID-range fix and comment correction landed, while the original automatic TOAST proposal and the later boundary-check and pinning ideas have no attached implementation or landing shown beyond the original TOAST patch.