[
https://issues.apache.org/jira/browse/KAFKA-20155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18060129#comment-18060129
]
Alieh Saeedi commented on KAFKA-20155:
--------------------------------------
[~mjsax] I think there is a way to optimize this so that we can avoid headers
deserialization for headers‑agnostic applications (which I assume is the
majority), but I agree that, in general, it’s not straightforward. We run into
similar challenges in the changelogging layer
(https://issues.apache.org/jira/browse/KAFKA-20179). We might instead state
that if a user wants to use headers‑aware stores, they need to accept the full
cost. Of course, we should provide benchmarking results so users can see the
overhead and make a more informed choice.
Regarding changing the byte format: should we really rule out lazy headers
deserialization forever, or keep it open so we can potentially add an
optimization later (implementing lazy deserialization for headers‑agnostic
apps)?
> Implement lazy deserialization mechanism
> ----------------------------------------
>
> Key: KAFKA-20155
> URL: https://issues.apache.org/jira/browse/KAFKA-20155
> Project: Kafka
> Issue Type: Sub-task
> Components: streams
> Reporter: TengYao Chi
> Assignee: TengYao Chi
> Priority: Major
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
