OK, I did not consider the case of the child process accessing the JCWs of a
parent process on a Linux machine.  There is definitely a different view
between Linux and MPE of child processes, which changes the related concepts
of JCWs and environmental variables which look very similar, but are thought
of and used in different ways.  Exposing  Linux and/or Posix concepts to MPE
just confuses the matter and adds ambiguity.  Although I covered the concept
before, I didn't explicitly cover JCWs.  I would prefer to store the JCWs in
a file (could be contained in a dynamic ram disc for I/O performance).  From
the Linux view the JCW file for JOE.SYS running session #S123 from MPE
started on May 16, 2006 at 9:30 am and 23 seconds would be
"~/SYS/users/JOE/jobs-2006-05-09-30-23/S123/.JCW".  From MPE-NG, everything
would be normal, and all child processes within session #S123 would "see"
the same JCWs.

As far as MPE files not looking like Linux files, it doesn't matter because
all of Linux is hidden behind an intrinsic wall.  The XFS filesystem has the
concept of Extended Attributes that are more than adequate for containing
MPE file labels and even have 3 name spaces to do fine grained protection.
The idea is to keep the MPE-NG world and the Linux/Unix/Posix world separate
to avoid complexity, confusion, and ambiguity.

BTW, most system tables would be contained (using the example above) in
"~/.boot-2006-05-09-30-23/" directory which could be a symbolic link to a
dynamic ram disc, or individual files could be memory mapped for
performance.

Once again, Linux/Posix/HFS is hidden behind a wall of intrinsics, and Linux
functionality is only exposed through that intrisic wall in a MPE-like way:
simply, securely, robustly, and with as much performance as possible within
budget (time/resources/money).

- Pete