Speculation either on the motives of those objecting to a release or of those making contributions or proposing a release does not advance progress. The accusations and counter-accusations seen on this thread are regrettable and I feel less and less confident in the future of Apache Hadoop as time goes on. As a strong believer in and advocate of open source as an answer to technical and architectural challenges, I am pained to see the members of what should be a vibrant community litigating in an ultimately self-defeating way. If only this energy put into argument could be channeled into code or patches...
In open source, if opinions were code we would rule the world. So what of this candidate? Artifact looks good, DFS tests are good, MR tests are good. Looked over some of the documentation and found no errors. To my knowledge this is now a superset of branch-0.20, addressing the reasonably determined deficit of rc0. There seems no reason other issues cannot be addressed subsequently. There has not been a release of Apache Hadoop 0.20 since at least Feb 6 2010 yet since this time important security enhancements have been contributed, but in the form of an Apache product these are only available as patches on a non-release branch. Forward progress of the Apache product seems more important than achieving the perfect release in all eyes. For example, append features remain on a non-release branch. I would really have liked to see the append changes included in this candidate, but this is not grounds for objection merely regret, and I hope this can be covered by a subsequent release, perhaps soon. After security and append features are in 0.20, in my personal humble opinion the 0.20 release in total is sufficient and all attention should be paid to the next release (0.22 or whatever), except for critical bug fixes. +1 Best regards, - Andy Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)