model-writtenrun 1076 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Thomas Munro proposed dropping MULE_INTERNAL because its obsolete implementation imposed maintenance and security costs, while leaving improvement as an alternative if actual users needed it #1.
- Tom Lane supported removal because multiple representations of the same characters undermine text equality, and Tatsuo Ishii agreed that the encoding had fulfilled its historical role, knew of no production users and clarified the Japanese character-set details #2 #3.
- Thomas submitted the removal patch, preserving the old encoding-number slot and renaming the surviving Cyrillic conversion module to cyrillic because it still handles four encodings #4.
- Jeff Davis found that upgrading a MULE_INTERNAL cluster failed during the upgrade itself and requested an explicit preliminary check; Tatsuo initially reported success after dropping his MULE_INTERNAL database, then agreed with Jeff after seeing a reproducer using a cluster initialized with that encoding #5 #6 #7 #8.
- Tatsuo found a remaining documentation reference to Mule internal code, and Thomas agreed to remove it before committing the original patch #9 #10.
- Thomas submitted a separate pg_upgrade check that validates every database encoding with PG_VALID_BE_ENCODING(), reports unsupported databases and recommends dumping and restoring with UTF8; Tom gave it a quick positive review #10 #11.
- Thomas committed the removal from #4 with only the extra documentation cleanup as 77645d44e3, then committed the unchanged upgrade check from #10 as ce11e63f81.
- Thomas confirmed that both changes were pushed, leaving no outstanding patch in the thread #12.