> although I don't see a particular difference The distinction I'd make us that a filesystem provides a common generic API that practically all processes on th

> although I don't see a particular difference The distinction I'd make us that a filesystem provides a common generic API that practically all processes on the OS understand and share access to. Pretty much always implemented out-of-process (in the kernel or another userland process via kernel stubs/hooks like FUSE). A data access library is usually much more specific to a particular data set or set of applications, and likely doesn't follow the filesystem abstraction (at least not in the same way).