The solution I posted only allows you to do anything once you chroot yourself, you get nothing by default. I think you can also run the directory layer in a mode in which subdirectories are prefixed with the root directory. My intention is actually that if you wanted the features of the directory layer (which you may or may not), you would re-implement it as a part of the application and the gateway wouldn’t have any knowledge of it. Your directory layer (as opposed to the global one used to enforce chrooting and where the authentication data is stored) is entirely yours.
You could imagine having language bindings that exposed the same API and almost the same semantics as the official bindings and porting code to use them would be trivial.