Hi @guozhangwang , @mjsax , and @bbejeck ,

I believe I've addressed all the comments thus far, with the exception of 
whether to store the values serialized.

I did notice a low-effort optimization in the current implementation to skip 
serializing if the buffer isn't size-constrained, which would carry over even 
when we add the changelog, if the buffer delays serializing until flush, so for 
high-turnover buffers, many records may never be serialized at all.

But that's all a little beside the point... I'm still happy to change the whole 
implementation to store serialized data if that's the reviewers' preference.

[ Full content available at: https://github.com/apache/kafka/pull/5693 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to