I noticed in the Flink 1.18 release that kafka connector is not following anymore the same version. Further the connector has now a separate repo/project on Github now. What were there reasons for this separation?

Almost all connectors are externalized and decoupled from Flink releases

thanks pointing to the wiki. I the reasons are great. https://cwiki.apache.org/confluence/display/FLINK/Externalized+Connector+development if somebody interested in details So If I understand it correctly for e.g. the kafka connector will keep the version

3.x.x-1.17

till flink 1.19 will be released and then the version

x.x.x-1.18

will be released for the kafka-connector?

So If I understand it correctly for e.g. the kafka connector will keep the version 3.x.x-1.17 till flink 1.19 will be released and then the version x.x.x-1.18 will be released for the kafka-connector?

Not per se, because the Flink Kafka connector itself can of course have a breaking change as well, so then it would be a new major or minor version. That depends on the maintainers. But there will be a x.x.x-1.18 version indeed, there's already a vote ongoing for that

Is there a recommendation how to upgrade flink version? For e.g. Flink 1.18 is out but the kafka-connector hasn't been released a version with/for 1.18. Does one should upgrade or should wait util all used connectors are ready for then next flink version?

It's not recommended to mix and match versions, but potentially it can work. If you want to be sure, you would wait for the officially released one

I'm not sure or whether its feasible to create flink bom, something like in the spring eco system where you as a developer can just reference a pom which handles all the dependencies. This might be also renovate/dependabot friendly.