I will try to explain my opinion on some of the questions being raised so far. Of course all these are open for discussion and nothing is final at this moment.
1. This code base is primarily targeted for usage by HBase and Scribe. Both of these are Apache open source projects. 2. Accordingly, my proposal is to name the branch hadoop-0.20-append. This name is more generic than the earlier one I suggested. 3. The proposal to pull patches into this branch will follow normal conventions that we follow in Hadoop. However, the release manager would have an option to veto a patch from being pulled into the branch. 4. if this type of branching causes too many forks of core Hadoop, then we can aim to merge some of them after some time. The time duration depends on the stability of the code in that particular branch and is difficult to predict. The PMC, of course, has to approve of any new branch; so, in effect can prevent undesired multiple forks if that becomes a problem in the future. 5. code changes to this new branch will go through the normal process via JIRAs, code reviews, unit tests, etc. 6. the goal is to have a standard hadoop release from this branch at some future point. if course, such a release has to be approved by the PMC. The release could be marked as "experimental" or some such thing if deemed appropriate by the PMC. thanks, dhruba On Sun, May 9, 2010 at 3:26 PM, Stack <st...@duboce.net> wrote: > On Sun, May 9, 2010 at 3:15 PM, Stack <st...@duboce.net> wrote: > > Should it be all of hadoop? Could it be hdfs only? > > > > Please ignore the above question (I just took a look at 0.20 repo). > St.Ack > -- Connect to me at http://www.facebook.com/dhruba