Standards Update, IEEE 1003.7: System Administration

[Jeffrey S. Haemer]

From: Jeffrey S. Haemer <jsh at usenix.org>


            An Update on UNIX* and C Standards Activities

                            December 1989

                 USENIX Standards Watchdog Committee

                   Jeffrey S. Haemer, Report Editor

IEEE 1003.7: System Administration Update

Steven J. McDowall <sjm at mca.mn.org> reports on the October 16-20, 1989
meeting in Brussels, Belgium:

Background

Joe Friday would say, "Just the facts, ma'am." And that's the way I
feel.  The facts are that I'm sick, it's Thanksgiving, I am going to
London for two weeks tomorrow, and 1003.7 is defining a standard way
to administer POSIX systems.

Now, almost everyone agrees that 1003.7 should deal with networks, not
just isolated systems.  To wit, it would be nice if I could administer
all the machines in a network from a single machine with simple
commands.  For example, to add a user to all machines in the domain
"mn.org", all I should need to do is issue a command like "adduser -d
mn.org -options -parameters username". The question is, without any de
facto standard already in place to adopt, how can we achieve this?

The Approach

This is important, so pay attention.  Because the major goal of 1003.7
is to create a standard way to manage a set of objects, the group has
decided to take an object-oriented approach.  Our idea is to begin by
creating a list of objects to manage, then to follow that by defining
the set of commands to manage each object.  This approach is novel for
both system administration and POSIX.  It will probably require more
work on the front end to define the objects, their attributes, and
their relationships, than to define the actual command structure to
support and manipulate them.  Whether this approach will work remains
to be seen.

__________

  * UNIX is a registered trademark of AT&T in the U.S. and other
    countries.

December 1989 Standards Update      IEEE 1003.7: System Administration


                                - 2 -

The Meeting

The meeting was boring.  To put it bluntly, the week was simply a work
week.  Objects (and sub-objects) were defined and discussed in detail,
then put in the draft.  Little got done on the first and last days,
due to EEC formalities, which left us with three working days instead
of the normal four and a half. Attendance was pretty dramatically
reduced, too.  About half the normal North Americans showed up,
probably because of the location, and only one (yes one...) new
European came even though we were meeting in Europe.  Oh well, except
for my having had my passport stolen, it was a good chance to see
Belgium.

Concerns

  1.  The process is taking a long time to move ahead, both because of
      the difficulty involved and because we seem to attract less
      manpower than many other groups.  Moreover, since we're taking a
      radical approach, it takes extra time to teach the ideas to
      anyone new that does come.

  2.  System administration doesn't have the glamour of some of the
      other areas being standardized.  As the Rodney Dangerfield of
      POSIX, 1003.7 gets no respect.

  3.  The notation we're using to define our objects is ASN.1.  "Why
      ASN.1?" you ask.  Simply because it's a standardized meta-
      language to describe abstract data types.  The feeling was that
      this would help make the whole package more suitable for
      interoperability.  I bring this up because there's some movement
      throughout 1003 to re-do all data structures in a new meta-
      language created by some of the people working on language-
      independence.  Not only would this require that we go back and
      re-do our definitions, but I also think ISO will only allow the
      use of standardized data-languages in their standards.  Does
      anyone out there know if there is such an ISO restriction?  If
      so, it's important for 1003 as a whole, not just for dot seven.

  4.  Currently, almost all working-committee members are from
      vendors.  IBM, DEC, HP, AT&T, and others are well-represented.
      A few interested parties, like OSF and /sys/admin.  are there as
      well, but as far as I can tell, there isn't one real user.  By
      "real user" I mean someone who does nothing but administer a
      system.  All of us are connected somehow with creating an
      administrable system or getting paid to do so.  Of course, I
      should make clear that we all have to administer systems of our
      own, so we're not simply an ivory tower group with no real

December 1989 Standards Update      IEEE 1003.7: System Administration


                                - 3 -

      experience, but representation is still grossly unbalanced.

  5.  Finally, there's been a loss of focus on interoperability
      directly attributable to the loss of our X/OPEN representative,
      Jim Oldroyd.  Jim was well respected and made many valuable
      contributions, but can no longer attend our meetings.  As the
      X/OPEN representative, he was very concerned with multi-vendor
      environments, and was a major force in helping us focus on and
      ensure interoperability.  I am not saying that no one else on
      the committee cares about the issue, but it does seem to be
      being pushed aside in a spirit of, "I think we shouldn't have
      any interoperability problems if we do this, so let's do it and
      worry about it later on." Jim had helped provide a more
      positive, direct approach of determining up front what would be
      needed for true interoperability. If X/OPEN is still interested
      in System Administration, and in making sure the 1003.7 standard
      includes provisions for interoperability, we could still use
      their help.

December 1989 Standards Update      IEEE 1003.7: System Administration


Volume-Number: Volume 18, Number 5