🚀 go-pugleaf

 
----------------------------------------------------------------------------
> Perhaps the better analogy for a large software
>system would be the space shuttle, which could only rely on a relatively
>_much_
>smaller set of existing components (Remember the Challenger?).
 
    I was the systems engineer on Shuttle's NSP (Network Signal Processor)
which, inspite of its name, was entirely dedicated hardware.  Like most space
qualified systems, the Shuttle's systems were deliberately kept as simple and
non-exotic as possible, using as few component types as they could get away
with.  It is presently flying on mostly 20-year-old hardware.
    The shuttle flies with 5 main computers, any three of which must be
operational.  They only recently replaced the core storage with semiconductor
memory.  One of the five computers runs on software that was written by a
totally different company that was kept isolated from the company writing the
primary software.  That was their ultimate insurance against software bugs.
All the hardware redundancy in the world won't help if all the software says to
land off the coast of Florida or in southern California's San Gabriels instead
of on the runway.
    The key to reliable subsystems is the use of standardized, well tested,
reliable components, and then thorough testing of the subsystem itself.  The
key to system reliability is to use only such sybsystems, do system analysis,
and provide system monitoring, redundancy, and thorough testing through the
most extreme conditions possible.
 
    Frankly, I personally don't like low-level programming (although I don't
denegrate those who love it).  I don't deal with the Mac Toolboxes any more
than I have to.  Sometimes there's no choice.  We need people working at all
levels.  However, I would like to see us move into providing reliable
subsystems that can be easily configured into powerful systems.  And many new
systems do not fit the classical (if the last decade can be called "classical")
concept of applications.
    We've escaped the shackles of the main-frame hardware, but we still have
ties to the "program" or "application" mentality.  IMHO, MacApp was a great
advancement but still was a remanent of that era.  As communications systems
and personal, portable computers become increasingly intertwined, standardized
reliable subsystems are going to become increasingly important.
 
G. Gordon Apple   D4887
Advanced Communications Engineering, Inc.
Redondo Beach, CA