Hi Ming-Yen chia_02: the example of exposing custom metrics should highlight that a third-party authorizer can have its own metrics. It doesn't need to follow the metric names used by StandardAuthorizer.
chia_03: please add something to "Rejected Alternatives". For instance, we could keep the unused method, but it could mislead third-party implementations. Best, Chia-Ping Ming-Yen Chung <[email protected]> 於 2026年9月19日週六 下午2:08寫道： > Hi Chia-Ping, > > chia_00: Yes. StandardAuthorizer is not a public API, so I updated the KIP > to remove StandardAuthorizer#aclCount() in 4.5. > > chia_01: Yes. I opened https://github.com/apache/kafka/pull/23521 for the > acls-total-count test and updated the Test Plan to reference it. It does > not depend on this KIP. > > Best, > Ming-Yen > > Chia-Ping Tsai <[email protected]> 於 2026年9月19日週六 上午1:50寫道： > > > Hi Ming-Yen > > > > chia_00: We could remove StandardAuthorizer#aclCount() without a breaking > > change, right? > > chia_01: It seems we could enhance the test cases for acls-total-count > > right now even if this KIP does not exist, right? > > > > Best, > > Chia-Ping > > > > Ming-Yen Chung <[email protected]> 於 2026年9月19日週六 上午1:37寫道： > > > > > Hi all, > > > > > > I would like to start a discussion on KIP-1342: Deprecate > > > Authorizer#aclCount <https://cwiki.apache.org/confluence/x/i5QmGQ> > > > > > > Authorizer#aclCount() was added for a broker-side metric that was never > > > implemented. StandardAuthorizer already publishes acls-total-count > > through > > > PluginMetrics, so this KIP deprecates the method in 4.5 and removes it > in > > > 5.0. > > > > > > > > > Best, > > > Ming-Yen > > > > > >
