Looks like it follows Amazon S3's design closely, rather than that of a classic distributed file system: * Immutable named objects * Addressable by chunk (eac

Looks like it follows Amazon S3's design closely, rather than that of a classic distributed file system: * Immutable named objects * Addressable by chunk (each object has a maximum size) * Optional per-object versioning * HTTP API In particular, while you could write a FUSE driver for it, it's designed for apps to use via an HTTP API, just like S3. It even implements the S3 API.