On 09/26/2011 09:08 AM, Roman Shaposhnik wrote: > Now that the work on incroparating the upcoming Hadoop 0.22 and > Hadoop 0.23 releases into Bigtop has started I've got a question > on what makes the most sense from a branching strategy point > of view. For example, it is pretty clear to me that the work > for 0.23 belongs in a separate branch in Bigtop. After all, > that code will be used no sooner than Bigtop 0.3.0. > > By that same logic one could say that the work on 0.22 also belongs > to a dediacted branch. I'm cool with that, but then I'm not quite > sure what status will trunk have. > > IOW, what is our policy to trunk commits? Do we only commit things > into the trunk that can rely on properly released apache projects, > or will it make sense for us to be a tad more aggressive and commit > things that we *hope* will make it to the next release (0.2.0 in this > particular case)? > > Thoughts? > > Thanks, > Roman.
My take on it is probably simple: * trunk should be in a releasable state * If there is some work which will break trunk or make it stable. Provided it is a finite amount of work with a known timeline for stabilization, I am ok to put it in trunk with optionally making a release just before introducing these changes * If there are more unknowns to the work, a branch would be more appropriate In this case, if we have some guarantees regarding hadoop-0.22 and the associated integration work, I would be ok for using trunk. But I would really prefer to get a release of bigtop out before that. Trunk has had a lot of improvements since 0.0.1 and we should make a release for hadoop 0.20.2 before jumping to the next version of hadoop. Thanks, Bruno
