I agree with Ted. It is important for Apache that every commit tie back to a public Jira. In the last month of Kylin's git log, I see a lot of instances of both no associated Jira and also umbrella jiras that have many commits. Using the git log for the last two months, grouping the 374 commits by the referenced jira you find only 31 jiras mentioned. KYLIN-650, which is mentioned the most often, has 16 commits. On the other hand, most of the commits are small.
I'd suggest a couple of things: * Rebase branches before merging them into trunk to make the change contained and easy to follow. * Always have a jira for every change and mention that the commit message * For large changes, break the work into parts and create a jira for each part. Jira has a mechanism to support that. Remember that with open source, you are trying to make your development process and history easy to follow. So making sure your git history isn't a complex tangle of git merges makes life better. Also being able to see what work people are doing and who did which work is critical. Thanks, Owen
