Cool, just created the feature branch "sentry-ha-redesign" @Colin: With respect to branch committers, Sentry does not have a branch committer role as of now. But, I can see how it can be useful to speed up branch development especially in the cases where the feature is completely being developed by non committers. Non committers can review and commit code to feature branch. And committers can review before the merge. Also, it would save committers review cycles if the branch never makes it to merge. But, we should note that there is a bigger penalty on committers at the time of merge.
Also doing so would involve adding feature branch committers to the ldapgroup ( I am not aware of other approaches if any), which is used in multiple places as a source of truth for committership in a project. Which can get messy quickly. An alternate approach would be to have "feature branch reviewers" +1 on patches on feature branch, but let one of the committers commit it. That way we have the similar advantages of "feature branch committers" but do not have to mess with ldap groups. Again, one of the committers should review before merge. I would propose going to the alternate route if community thinks it would be useful. I can start a new discussion thread about this. For now, we can get started using the feature branch "sentry-ha-redesign" as is without being blocked on this decision. On Thu, Jun 9, 2016 at 4:45 PM, Colin McCabe <cmcc...@apache.org> wrote: > Thanks, Sravya. That seems very similar to the way I've seen feature > branches work in Apache Hadoop. > > Hadoop has an additional role called "branch committer." See > https://hadoop.apache.org/bylaws.html and search for "branch committer." > Basically, branch committers are authorized to make commits only to a > specific git branch, not to any other branches. The goal of adding the > branch committer role was to make feature development faster by allowing > patches for a specific feature to be reviewed and committed by a larger > group of people. Of course, new features are reviewed before merging to > master. Sentry might want to think about adding a role like this, if it > seems to make sense. > > best, > ColinM > > > On Thu, Jun 9, 2016, at 14:36, Sravya Tirukkovalur wrote: > > Cool. > > > > @Colin: Let me explain how feature branches worked so far in the Sentry > > community. And I would love to here what other communities do, and if we > > can adopt some of the best practices. > > - Committing: We create sub tasks and each sub task is reviewed as usual > > and is committed by one of the committers after a +1 from committer. > > - Merging: We ask the feature owner to rebase on master. And one of the > > committers applies these commits on to the master. Or, we merge it with > > master. > > > > Thanks, > > > > > > > > > > On Thu, Jun 9, 2016 at 2:21 PM, Hao Hao <hao....@cloudera.com> wrote: > > > > > +1 for having a feature branch. Thanks! > > > > > > Best, > > > Hao > > > > > > On Thu, Jun 9, 2016 at 1:30 PM, Colin McCabe <cmcc...@apache.org> > wrote: > > > > > > > By the way, this is the first community I've been in that has two > > > > Colins! > > > > Hope it's not confusing :) > > > > > > > > > > > > > > > > > > > > On Thu, Jun 9, 2016, at 13:29, Colin McCabe wrote: > > > > > Thanks, Sravya. > > > > > > > > > > +1 for a feature branch, so that we don't have to worry about > things > > > > > like what release it should appear in until the feature is > complete. > > > > > > > > > > I am interested in working on this feature, and I think Rahul > Sharma is > > > > > as well (both from Cloudera). If we decide to do a feature > branch, I > > > > > would be interested in being a branch committer and I believe Rahul > > > > > would as well. > > > > > > > > > > cheers, > > > > > Colin > > > > > > > > > > > > > > > On Thu, Jun 9, 2016, at 12:21, Sravya Tirukkovalur wrote: > > > > > > Hi folks, > > > > > > > > > > > > Shall we create a feature branch for Sentry HA redesign or do > you all > > > > > > feel > > > > > > comfortable committing it to master? > > > > > > > > > > > > Thanks! > > > > > > > > > > > > > > > -- > > Sravya Tirukkovalur > -- Sravya Tirukkovalur
