Hey Mikhail, I believe we may have a similar concerns supporting older versions of HBase but I think it is almost always better to fix the problem on trunk branch first, and then backport the patch to an older version that you need to specifically support. (such as 0.90.x in my case or in your case 0.89fb).
I know this is a burden, but I think it may works out best in the end. If done trunk first, everyone in the community will have a stake in the patch and you'll get likely get more folks to review the work. Doing this will better guarantee that forward progress goes into trunk. At the end of the day, many of the committers are focused on the newer versions and would like to focus there more. Jon. On Mon, Feb 6, 2012 at 11:36 AM, Mikhail Bautin < [email protected]> wrote: > Hello Everyone, > > Some of you have probably been wondering about what these "[89-fb]" patches > that our team submits for review are, so I would like to clarify that a > little bit. We run a custom version of HBase based on 0.89 at Facebook, > codenamed "0.89-fb", but we do our best effort to submit all of our > improvements to the trunk as well. As a result, we frequently put an 89-fb > version of a patch for review first, go through a review loop, and only > then put the trunk patch out for review. We have noticed that in such > situations our trunk patches sometimes receive many more comments than the > earlier 89-fb versions of the same patches, which complicates our > development workflow, because we have to go back and make these additional > changes as a follow-up patch to 89-fb. > > It would greatly simplify our workflow if people treated 89-fb patches just > like any other patches, and submitted most of their feedback on our code > contributions (consisting of an 89-fb patch and a trunk patch) as part of > whatever patch is published first. In other words, I would like to ask you > to treat 89-fb patches just the same as trunk patches, because a trunk > patch is likely to follow. That was our hope when we open-sourced our > internal version of HBase and moved our code review workflow to the > externally-visible review system at http://reviews.facebook.net. The only > kind of 89-fb patches that we are not planning to port to trunk are tagged > [master], containing custom changes to the 89-fb master code. > > It would be great to hear what you think about the above and how we can > make it easier for you to give us early feedback on our code contributions. > > Thank you! > --Mikhail > -- // Jonathan Hsieh (shay) // Software Engineer, Cloudera // [email protected]
