[GitHub] [kafka] ijuma commented on pull request #9689: KAFKA-10740; Replace OffsetsForLeaderEpochRequest.PartitionData with automated protocol

2021-01-22 Thread GitBox
ijuma commented on pull request #9689: URL: https://github.com/apache/kafka/pull/9689#issuecomment-765497872 It definitely requires analysis on a case by case basis. However, having helpers that do conversion lazily also have performance issues if the method gets called repeatedly, unless

[GitHub] [kafka] ijuma commented on pull request #9689: KAFKA-10740; Replace OffsetsForLeaderEpochRequest.PartitionData with automated protocol

2021-01-22 Thread GitBox
ijuma commented on pull request #9689: URL: https://github.com/apache/kafka/pull/9689#issuecomment-765474499 It's the nature of protocol evolution. You don't know how things will change over time ahead of time. This is an au

[GitHub] [kafka] ijuma commented on pull request #9689: KAFKA-10740; Replace OffsetsForLeaderEpochRequest.PartitionData with automated protocol

2021-01-22 Thread GitBox
ijuma commented on pull request #9689: URL: https://github.com/apache/kafka/pull/9689#issuecomment-765233789 Like Jason said, I think the main challenge is how to deal with protocol evolution. It's typically possible to deal with, but it can require quite a lot of contortion. LeaderAndIsrR