Hi Ethan, This proposal sounds awesome!
- There are significant feature gaps between Hudi 0.x and master branch already. These gaps would make migrating Hudi users to the shiny 1.0 release extra challenging. - Certain features like Hadoop-agnostic abstractions and Trino native connector are essential to AWS Athena/EMR Hudi + Trino users. Migrating to a version with these features is not optional but a necessity for them. Keeping 0.x releases going for a while can help ease the slope and boost users' confidence in adopting new features. Eventually, this can help more people migrate to Hudi 1.x. Best, Shawn On Thu, Feb 15, 2024 at 12:25 AM Y Ethan Guo <yi...@apache.org> wrote: > Hi folks, > > As the community is preparing 1.0 GA release, I feel it is still valuable > to have another release in 0.x series, i.e., 0.15.0 release, to support > better engine integration, including Spark 3.5 support, Flink 1.18 support, > Trino Hudi native connector support with Hadoop-agnostic abstractions and > APIs. I volunteer to be the release manager of the 0.15.0 release, given > that I drive or am familiar with these changes. > > To maintain the stability of 0.x releases, I propose that we should mostly > merge bug fixes and changes resolving regressions to 0.x releases. No > backward incompatible format change should be included in future 0.x > releases. Any improvements or major features should be excluded except for > rare occasions at RM's discretion. > > As we advance, I also propose that we have a process to merge changes from > master to branch-0.x [1] when necessary. From now on, the author of the > original PR merged to master is responsible for raising another PR against > branch-0.x, if the author wants the change to be in 0.x releases. > > Let me know your thoughts. > > [1] https://github.com/apache/hudi/commits/branch-0.x/ > > Thanks, > - Ethan >
