Re: What can a translator do that FUSE can’t?

[Roland McGrath]

There are two core things about translators vs other systems' filesystems:

1. passive translators.  Other systems approximate this with things like
   automount mount points.  Hurd translators can be (and usually are)
   permanently associated with a file on the containing filesystem.  In
   Unix terms, it means that every file might be an automounting mount
   point.  There is no central administrative overhead associated with
   that, so it can be used more casually and pervasively than filesystem
   mountpoints traditionally are, even with FUSE.

2. They are naming points for arbitrary RPCs.

   In FUSE, the only kind of interface available is the filesystem
   interface.  If you want to have special magic operations, you represent
   those in the vocabulary of a filesystem.

   The Hurd is a generically RPC-based system in a deep way.  Every kind
   of subsystem in the Hurd universe is contacted via RPCs, with
   appropriately specialized calls for whatever you want to do.
   Filesystems (translators) are the standard rendezvous point for
   finding a receiver for RPCs, but they do not constrain the interface
   that client and server can use once they've made the rendezvous.


Thanks,
Roland