On 05/05/11 18:52, Todd Lipcon wrote:
On Thu, May 5, 2011 at 10:32 AM, Eric Yang<[email protected]> wrote:
Git is powerful in maintaining different branch of the source code.
However, it will only work if the entire community is willing to move to
git. Maintaining svn and git hybrid, is a time consuming task that we are
paying in full price. Hadoop community should work smarter for the source
control. What do people think about fully adopting git instead of svn?
+1 for Git as a tool. But using git makes it even _more_ important that we
have a clearly defined release process that outlines which branches are
meant to be released as official artifacts, and what the inclusion criteria
for those branches should be.
I'm +0.9995 for git: some bits I like, some bits I don't (it's awful for
binary data). And you need more than just a release process locked down,
you need the developers understanding following a good process. If you
have written down process docs there, I'd love to see them.
apache infrastructure are discussing git -what would be best would be to
start with a non-critical project, such as one or more of the moved
contrib projects (like MR-Unit), so we can see that it, gerrit, etc work
well within the Hadoop developer world.
