RFS vs. NFS

Dick St.Peters stpeters at dawn.steinmetz
Sun Apr 3 08:53:15 AEST 1988


In article <10186 at ulysses.homer.nj.att.com> ekrell at hector (Eduardo Krell) writes:
>You're completely missing the point. When I have a network of Suns,
>Vaxen and other boxes running UNIX, I DEFINITELY WANT UNIX file
>system semantics on remote files. I don't want my programs to
>be aware of the fact that the file they're operating on is remote.
>This is what is meant by UNIX file system semantics: the behavior
>is the same whether the file is local or not. It's called transparency.

It seems to me that "UNIX file system semantics" is not well-defined
in a network environment, particularly if transparency is the goal.

Consider RTI's Freedomnet, which RTI advertises as supporting full
UNIX file system semantics.  However, if program foo physically
resides on a remote machine, typing foo causes foo to run on the
remote machine.

That's not what one would normally expect, but it happens to be very
convenient when foo is, say, a VAX binary executeable resident on a
VAX and you happen to give the command on a Sun.  For most programs,
this is about as transparent as you can get: from a computation (data
in --> data out) point of view, you get the same result whether the
file is local or not, thus meeting ekrell's criterion.

However, if instead of foo, the program is reboot ... well, you get
the idea.  That's a pretty pathological example, but if you're running
on a Cray and happen to type a command resident on a PC, you might
feel that the extreme transparency of Freedomnet isn't what you want.

Even under RFS, exec'ing a VAX executeable on a Sun doesn't work very
well, so it can matter whether the file is local or remote.  RFS thus
does not support full UNIX file system semantics according to the
definition ekrell gave.

It appears that "transparent" and "behaving the same whether remote or
local" are not equivalent, and further, neither is adequate as a
criterion for the meaning of "supports UNIX file system semantics".

I have my doubts whether there is any adequate definition of UNIX file
system semantics for a heterogeneous distributed environment.
--
Dick St.Peters                        
GE Corporate R&D, Schenectady, NY
stpeters at ge-crd.arpa              
uunet!steinmetz!stpeters



More information about the Comp.unix.wizards mailing list