Thanks David. This is a nice improvement and I am in favour. I don't think this is a controversial change, so my points are minor:
chia1: We should keep in mind that this could eventually be renamed to just "memberEpoch," and every rename requires actual changes in the user code (as opposed to renaming RPCs), so this could factor into whether we need the intermediate "generationIdOrMemberEpoch". Not strictly against it though. LB1: Not strictly required, but consider specifying the error order. In particular, if a topic does not exist, does the RPC return UNKNOWN_TOPIC_ID or UNKNOWN_TOPIC_OR_PARTITION? Based on the OffsetCommit v10 handler, I'd expect the former, but making it explicit would be beneficial. Cheers, Lucas On Thu, Apr 23, 2026 at 1:46 PM Lianet Magrans <[email protected]> wrote: > > Thanks for the KIP David, nice to see this alignment! > > LM1: About topic ID resolution & adding the topics to the producer > metadata. The topics used for the sendOffsetsToTransaction will > commonly not be the ones found in the producer metadata (topics being > produced to), so agree with the waiting to resolve, otherwise seems we > wouldn't hit the topic ID path, but should we keep them around in > metadata, or treat them as "transient" topics (removed from metadata > after the call completes, similar to what we have in the consumer > path). I imagine the latter, mostly to avoid polluting the producer > metadata, but thoughts? > > Agree with chia1 btw (we already have confusing mapping for that > generation Id on the client side, to epoch in some places, to > generation in others) > > Best, > Lianet > > > On Wed, Apr 22, 2026 at 2:19 PM Chia-Ping Tsai <[email protected]> wrote: > > > > hi David > > > > the KIP is excellent. I have only one small question. > > > > chia1: Regarding the renaming of `GenerationId` to > > `GenerationIdOrMemberEpoch`: I'm wondering if we should also align the > > naming of `ConsumerGroupMetadata#generationId` to stay consistent? > > > > Best, > > Chia-Ping > > > > Andrew Schofield <[email protected]> 於 2026年4月22日週三 下午3:32寫道： > > > > > Hi David, > > > Thanks for the KIP. Just one query. > > > > > > AS1: Is there any reason why GenerationIdOrMemberEpoch is ignorable in > > > OffsetCommitRequest but not in TxnOffsetCommitRequest? > > > > > > Thanks, > > > Andrew > > > > > > On 2026/04/21 20:13:20 David Jacot wrote: > > > > Hi, > > > > > > > > I would like to start the discussion on KIP-1319. This is a minor KIP > > > > which brings the TxnOffsetCommit API in line with the OffsetCommit > > > > API. > > > > > > > > https://urldefense.com/v3/__https://cwiki.apache.org/confluence/x/NJI8G__;!!Ayb5sqE7!tu8P8SATZdib2PXwyaGOjmhzmuTByIKc4RunefP3j2stfdWC2eB00wOgGLtgJHiWdP1QNyofmPZdQbCUXr5BVQ$ > > > > > > > > Thanks, > > > > David > > > > > > >
