Hi,

I agree that releasing a new connector for every minor Flink release might not 
be strictly necessary. However, as long as the Flink version is embedded in the 
connector's versioning, it creates a direct logical link between the two.

If our goal is to avoid new releases as long as compatibility remains intact 
(provided APIs don’t change, as Martijn suggested), I believe we should 
decouple the Flink version from the connector's version and maintain a 
compatibility matrix in the documentation instead.

Under this approach, for every new Flink release, we would simply update our 
CI/CD to verify the new version and update the matrix in docs. And this will 
also allows us to release a single version of the connector rather than 
managing multiple releases for every Flink version we wish to support.

Ultimately, I think it comes down to where we want this 'matrix' to live: in 
the code/artifacts or in the documentation. I understand that a connector 
version like X.Y.Z-2.0 might technically work with Flink 2.2, but without an 
updated version or a clear matrix, we are essentially asking the user to assume 
that risk themselves.


Regarding the JDBC connector, Ferenc, I’d be happy to help out with that. I'm a 
regular contributor to that repository.

Best regards,
Boto

Reply via email to