The readme file from the actual implementation can be found in README.UNFS
Performace of the server is about 25% of a kernel based NFS running on the same machine. The performance of unfsc (the RPC shuffler) is even worse as it has to unpack/pack each RPC twice.
Because the server can be configured to re-export NFS mounted filesystems plus the fact that by default the server ignores non-NFS mount points, only a single mount per server need be issued on the client to access all of the server's filesystems.
The following entry needs to be added to /etc/services
unfs 449/tcpAnd the following in /etc/inetd.conf
unfs stream tcp nowait root \ /usr/local/etc/unfsd unfsd-tcp -aof course the portnumber used can be changed to any available number. The -a option tells unfsd to require a one-time password before processing the mount request. This prevents auto-recovery if the server is lost - a real pain.
For the next release (please don't ask when :-) I plan to support RPC's over SSL and use X.509 certs to authenticate mount requests as an option to one-time passwords. This would allow for auto-recovery from server restarts and would of course also allow for safe export of filesystems beyond the firewall.