RFS vs. NFS

[Skip Egdorf]

Please note that comp.arch has been removed from the Newsgroups line.
This is no longer a computer architecture issue. PLEASE be good
net.people and acquire the practice of at least looking at who
will receive your followup.

In article <4188 at chinet.UUCP>, les at chinet.UUCP (Leslie Mikesell) writes:
> In article <7556 at brl-smoke.ARPA> gwyn at brl.arpa (Doug Gwyn (VLD/VMB) <gwyn>) writes:
>
> >I agree with your comments, but to be fair it should be noted
> >that one of the explicit design goals of NFS was to work not
> >only with UNIX filesystems but also with MS-DOS filesystems.
>
> AT&T sells a DOS server product for the 3B line that provides
> netbios compatible file and print services to PC's connected
> via the Starlan network.

Recently, some users of a large AI application on a Symbolics 3600
felt the need to produce some pictures with Microsoft Chart.
The Symbolics, for good reason, is a standalone machine that
is not allowed to participate in our general network. The users
spent a day trying to figure out how to get the data from the
Symbolics to the PC. Suggestions were made regarding serial lines,
translating Kermit to Lisp, and other worthwhile ideas. I plugged
a 3-com card into the PC, strung a local ethernet to connect the two
machines, loaded Sun's PC-NFS onto the the PC and ILA's EXCELLENT
Symbolics implementation onto the 3600, and had the machines talking.
The Symbolics performed as the NFS server and the PC mounted
	lispm:> as D:

The users then produced their charts using the seemingly local
MS-DOS files. Nary a Sun in sight!!

ALL THE WORLD'S NOT UNIX.

I was particularly struck at the Atlanta Usenix a few years back
that the Remote File System issue was really getting shaken out.
Everybody was giving papers on their remote files systems.
Stateful vs Stateless was suddenly a hot topic. AT&T was the
ONLY vendor pushing "Pure Unix Semantics." I was puzzled when
no one else seemed to realise that this is because AT&T was
now a REAL COMPUTER COMPANY whose only goal in life is to lock
you into AT&T's product. (Ok, so they're not very good at it!)
Naturally RFS preserves full Unix semantics. Why would AT&T
want to encourage you to use all those VMS Vaxen, Lisp Machines,
PCs, when they can lock you in with FULL UNIX SEMANTICS?

I will never again be tied to a single vendor. I saw too many
of my DECsystem-10 friends left twisting in the wind by DEC
to let that happen. I really like Unix. I really like my Sun.
However,

Note that the Symbolics NFS is an independent implementation
based on the NFS specs placed in the public domain by Sun.
NFS is the closest thing that I can find to a vendor independent
yet available remote file system. RFS doesn't come close on
either count.
NFS needs some things to be really vendor independent (e.g. support
for version numbers on VMS and Lisp Machines would be nice)
However, they can be added to the protocol. RFS cannot ever
support such things.

					Skip Egdorf
					hwe at lanl.gov

Usually I don't include disclaimers, but this time:

Dear AT&T lawyers: These are my opinions, Not Los Alamos
National Laboratory's. Note that this is written on a Sunday
Night from a terminal at my home. Please don't take away
all the Lab's nice Unix licenses. Thank you. Have a nice day.