[267] in Commercialization & Privatization of the Internet
Re: Alternatives to NFS (Was: Re: Should the NREN be funded?)
daemon@ATHENA.MIT.EDU (Sean McLinden)
Mon Mar 4 09:40:18 1991
Date: Mon, 4 Mar 91 09:18:26 -0500
From: sean@dsl.pitt.edu (Sean McLinden)
To: david@TWG.COM, thakur@zerkalo.harvard.edu
Cc: com-priv@psi.com
While it is true that there are technical difficulties with large
scale (campus wide or greater) NFS implementations, it is a bit unfair
to say that AFS is an alternative. The two, after all, were designed to
do different things and their designs were rooted in different models
of the world. Even now there are things for which I would not use
AFS just as their are environments where I would not consider NFS.
As someone mentioned, there are other efforts as well: RFS, Vax
Cluster, and many fine experimental systems which offer advantages
over both AFS and NFS.
Most importantly, however, is the simple fact that there is always
a limit to what you can do at the level of the filesystem in *any*
distributed system as long as the filesystem is viewed as, essentially,
a static structure into which are fit little pockets of dynamic data
(files).
One of the interesting applications of such systems as MACH is the
thought that a filesystem could be a process which, in turn, could
be a database that could be mapped into a static Unix-like file
system at the level of library routines. One could imagine all sorts
of applications of this.
I guess the point is that the discussion gets a little far afield if
we start talking NETNEWS vs. LISTSERV or NFS vs. AFS in the context
of NREN. It's not the protocol that describes the need, it is the
application.
Sean McLinden
Decision Systems Laboratory
University of Pittsburgh Medical Center
