DonHopkins
7 years ago
0
0
Here's some old but interesting discussion about NeFS that I saved from Comp.protocols.NFS. (they were SHOCKED I say SHOCKED!!!)

http://donhopkins.com/home/archive/NeWS/Comp.protocols.nfs.N...

    From: brent@terra.Sun.COM (Brent Callaghan)
    Newsgroups: comp.protocols.nfs
    Subject: NeFS Protocol Spec Available
    Keywords: NFS version 3
    Date: 13 Feb 90 01:57:31 GMT

    NeFS is not NFS.  It started out as an NFS protocol rev
    (NFS version 3) but its protocol is a radical departure
    from the NFS's remote procedure call model.  In order
    to avoid confusion with NFS we're calling it NeFS for
    now - Network Extensible File System (the similarity of
    the name to a window system protocol is entirely intentional).

    A draft spec for this new protocol is available via anonymous
    FTP from titan.rice.edu (128.42.1.30) as

        /public/nefs.doc.ps

    The file is ~250K of PostScript.  It prints 52 pages.
    The spec is not available in any other format.

    Comments, suggestions, flames etc are welcomed.  Direct them
    to "nfs3@sun.COM".

        Thanks

    Made in New Zealand -->  Brent Callaghan  @ Sun Microsystems
                 uucp: sun!bcallaghan
                 phone: (415) 336 1051


    From: mo@messy.bellcore.com (Michael O'Dell)
    Newsgroups: comp.protocols.nfs
    Subject: RE: NeFS protocol
    Date: 23 May 90 11:57:34 GMT
    Organization: Center for Chaotic Repeatabilty

    Dave Clark, Internet Architect, just gave a talk here at Bellcore and
    he was lobbying for exactly the kind of approach being proposed by the
    new NeFS protocol - don't send packets, send programs!

    I do recommend you read the NeFS document and think about it hard
    before you jump to any conclusions.  The proposed model is novel, and
    goes a long way toward moving us to "action at a distance" instead of
    "remoting local operations."  This turns out to be vitally important if
    networks are to scale with latency, since, as Dave Clark so eloquently
    showed, for large networks (ie, country-sized), as bandwidth goes to
    infinity latency goes to 30milliseconds, and you can't do anything
    about that, because you can't change the speed of light.  So, to do
    very well in the face of that, you must avoid round-trip delays like
    the plague.  Sending a program to the server can save many round-trips
    for many operations.

    Anyway, it is interesting reading.  Particularly in light of Dave
    Clark's talk.

        -Mike