[ https://issues.apache.org/jira/browse/KAFKA-9559?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17314586#comment-17314586 ]
Matthias J. Sax commented on KAFKA-9559: ---------------------------------------- {quote}refactoring the code where we get the default serde and maybe/maybe not use it? {quote} Yes, that was my proposal, ie, only try to get the default serde if we really want to use it. {quote}Was there a reason you preferred the null check over the poison serde? {quote} IMHO, it would add an (unnecessary?) layer of indirection and thus making the code more complex? To me, it seems cleaner for follow the refactoring idea – it's of course only my personal opinion. > Change the default "default serde" from ByteArraySerde to null > -------------------------------------------------------------- > > Key: KAFKA-9559 > URL: https://issues.apache.org/jira/browse/KAFKA-9559 > Project: Kafka > Issue Type: Improvement > Components: streams > Reporter: A. Sophie Blee-Goldman > Assignee: Leah Thomas > Priority: Blocker > Fix For: 3.0.0 > > > The current default "default serde" is not particularly useful, and in almost > all cases is intended to be overwritten either by setting the config > explicitly or by specifying serdes directly in the topology. If a user does > not set the config and misses specifying a serde they will get a runtime > ClassCastException once they start processing unless they are in fact > processing only bytes. > We should change the default default to null, so that an exception will > instead be thrown immediately on startup if a user failed to specify a serde > somewhere in their topology and it falls back to the unset default. > -- This message was sent by Atlassian Jira (v8.3.4#803005)