@nacx Note that the changes are for the exception only case. This means that the client will send a request but won't receive the response. It doesn't matter whether the provider uses a proper REST API or not - if an exception occurs in the middle of a request, then it's not clear whether it can be retried safely. All providers could end up in the same scenario - retry a create machine call and get a leaked machine as a result.
I don't mind doing it for SL only, but looks like all providers could benefit. wdyt? --- Reply to this email directly or view it on GitHub: https://github.com/jclouds/jclouds/pull/744#issuecomment-101248687
