Also, committing directly to the ASF git repo circumvents the CI jobs when a pull request is made in GitHub. Those jobs could catch any obvious problems and provide a layer of protection from breaking the build.
What's unfortunate in this case is that we *also* have CI infrastructure that is supposed to flag up changes by commits such as this, which would allow us to quickly revert them if necessary.
It looks like the Apache Nexus server has been throwing 503s all over our builds [1] :-( I hope we'll be back to normal soon.
ap [1] https://jclouds.ci.cloudbees.com/job/jclouds/2637/console
