(was slashes, now NFS devices)

[Frank Peters]

In article <14363 at ulysses.att.com> ekrell at ulysses.att.com (Eduardo Krell) writes:

> In article <BZS.91Feb22163556 at world.std.com> bzs at world.std.com (Barry Shein) writes:

> [about NFS special files]

>>Obviously they should be interpreted
>>relative to the local host only.

> Not so obvious to me. After all, the server is probably the only machine
> which can interpret the meaning of that special file in a sensible way.
> By forcing a local interpretation of special files, you lose the ability
> to access remote devices.

> If you can access remote files, why can't you access remote devices using
> the same mechanism? Under RFS, special files are interpreted by the server.

You ignore (or aren't aware of) two important differences between RFS
and NFS that make this impractical.

1. NFS is designed to be operating system independant while RFS
   assumes UNIX on both ends of the connection.  The concept of a file
   system entry that points to a device by magic major and minor codes
   is very OS specific.  While that concept is by no means unique to
   UNIX it is far from universal and implemented differently wherever
   it is found.

2. This is probably most important.  NFS is stateless. Whether this
   statelessnes is a good thing or not is another debate.  But it
   means that I can open a file on a remote host via NFS and be in the
   middle of updating it when the server reboots and when the server
   comes back up the operation will continue.  Mapping this
   functionality to tape devices would require the server to
   assure that the tape is positioned to EXACTLY the position it was
   at when the reboot occured so that the remote tar can continue
   unabated.  Even if this were practical it would be a royal pain.

No, I think special files must be treated like any binary data.  If I
mount my sparc executables on a MIPS box those files are meaningless
data.  If I want to right a sparc simulator I should be able to access
those files as executables for my simulator.  I shouldn't and can't
execute that sparc binary and make it run on the server.  In the same
way, a special file mounted on a non UNIX box should be meaningless
data.  If I choose to write a UNIX device file interface emulator for
VMS or DOS I should be able to pass it the special file data for local
interpretation.  I shouldn't be able to access those files and have it
affect the remote server.

Frank
--
Frank Peters   Internet:  fwp1 at CC.MsState.Edu         Bitnet:  FWP1 at MsState
               Phone:     (601)325-2942               FAX:     (601)325-8921