model-writtenrun 7371 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Jeff Davis opened the thread in #1 because SQL slot creation supported two-phase decoding while CREATE_REPLICATION_SLOT did not, despite an unused two_phase field in its command structure.
- Amit Kapila explained in #2 that protocol support had awaited subscriber-side work and testing, while Jeff advocated exposing it through pg_recvlogical in #3; Amit proposed a separate cleanup followed by protocol and client support in #4.
- Ajin Cherian supplied the cleanup and initial protocol patches in #5, and Amit committed the cleanup unchanged as be57f21650, confirming the push in #7.
- Ajin added pg_recvlogical support in #8; Jeff tested it and requested keyword placement, protocol documentation, option validation, and a physical-slot assertion in #9, which Ajin incorporated in #10.
- Amit proposed using the existing slot-option list, adding TAP tests, and checking server versions in #11 and #12; Jeff raised grammar-level separation of physical and logical options in #13, while Tom Lane favored downstream validation for clearer errors in #15.
- Amit explained in #16 that subscriber-side enabling after initial table synchronization could remain separate, and Ajin's revision in #17 adopted option-list processing, backend validation, server-version gating, and tests.
- Amit reviewed the revised patches as mostly ready in #18, listing cosmetic fixes and considering PostgreSQL 14 inclusion, while Ajin supported exposing protocol parity and deferring subscriber changes in #19.
- Jeff expressed concern about spreading protocol changes across releases in #20, leading Amit to choose the PostgreSQL 15 development cycle in #21; Ajin addressed cosmetic feedback in #22.
- At the thread's last message, #23, Ajin reported incorporating the remaining patches into the larger subscriber-side two-phase series elsewhere, so the cleanup had landed while the feature had moved onward for PostgreSQL 15.
- The supplied later commit cda03cfed6 landed both feature patches from #22 with light cleanup, documentation edits, and PostgreSQL 15 version gating; subscriber-side START_REPLICATION changes remained outside this commit.