model-writtenrun 8398 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed making SCRAM the default password encryption to encourage adoption while leaving the source-build authentication default unchanged in #1; Tom Lane asked about client readiness in #2, and Magnus Hagander reported that nearly all listed drivers supported SCRAM in #3.
- Stephen Frost favored changing defaults during the PostgreSQL 13 beta in #6, but Tom Lane argued that this could invalidate beta testing in #7; Jonathan S. Katz supported targeting PostgreSQL 14 in #11, with Tom clarifying that PostgreSQL 9.6's approaching end of support strengthened that timing argument in #13.
- Peter posted the PostgreSQL 14 patch in #15, changing backend and configuration defaults to SCRAM and removing initdb's special override; Jonathan's review in #16 prompted the revision in #17, which preserves explicit MD5 selection unless SCRAM is also requested.
- Michael Paquier questioned the retained Boolean-style aliases for MD5 in #18; Peter proposed removing them in #19, and Magnus and Michael supported removal rather than retaining confusing legacy mappings in #20 and #21.
- Jonathan tested the revised initdb authentication combinations and identified confusing missing-password diagnostics in #22; Peter proposed a single method-independent error message in #25, which Jonathan accepted in #27.
- Peter also proposed removing the ENCRYPTED and UNENCRYPTED role keywords in #26, prompting Robert Haas and Stephen to question implicit password-validator detection and storage design in #28 and #29.
- Michael favored retaining ENCRYPTED for compatibility in #31, while Jonathan treated keyword removal as a separate patch in #32; Stephen opposed indefinite deprecation in #34, and Tom supported removing the keywords for PostgreSQL 14 in #36.
- Peter committed c7eab0e97e from #17, adding legacy-alias removal, corresponding documentation and regression changes, and the simplified initdb diagnostic, then announced completion in #37.
- The thread ended with acknowledgments from Jonathan in #38 and Michael in #39; the password-encryption change had landed, while the separate role-keyword removal and validator-design proposals had no attached implementation or landing shown here.