Currently S3Proxy uses jclouds for the transient (in-memory) and filesystem (on-disk) providers for local storage. Given jclouds declining activity, it would be prudent to bring more of its functionality into S3Proxy itself which would simplify debugging and releasing. Implementing a nio2 filesystem backend would allow not only local on-disk storage, but also memory via jimfs, and potentially crazier things like Hadoop. References #606.