Conference 7.286::space

Date: 31 Oct 84 08:01:11 PST (Wed)
To: Ross Finlayson <rsf@su-pescadero>
cc: space@mit-mc
Subject: NASA and space weapons
From: Martin D. Katz <katz@uci-750a>

   NASA does NOT (and I hope, will not) develop space weapons.
   NASA is a civilian agency.

The fact that NASA is a civilian agency does not mean that they do not do
research and development on weapons systems.  NASA does not build space
weapons, but they do deploy them (deliver them to orbit).  In addition, NASA
is responsible for the research leading to development of boosters, advanced
aircraft (e.g. "Stealth"), etc.  All such research performed by military
agencies is coordinated with NASA.  In addition, any major increase in
military payload to orbit will mean more money to NASA (unless it all goes
via military boosters).

    Remember skylab? What a weapon if it had hit something important.
    All that comes down is not always nuclear. A 1 kilometer asteroid would
    do more damage than a nuclear attack. The first nation to the asteroid
    belt wins.

    	See SPACE Note 165.
    
    	Larry
    

    Costly Bugs: As complexity rises, tiny flaws in software pose a      
        growing threat. Defects have killed sailors, maimed sick, hurt
        firms; 

        The shuttle-bus solution 
	Is 'Star Wars' too vulnerable?

    [What follows are extracts, equal to about 1/3 of the article - TT]

     During the past five years, software defects have killed sailors,
  maimed patients, wounded corporations, and threatened to cause the 
  government-securities market to collapse. Such problems are likely to
  grow as industry and the military increasingly rely on software to
  run systems of  phenomenal complexity, including President Reagan's
  proposed "Star Wars" anti-missle system. "The vast majority of
  systems are deeply flawed from the  viewpoint of reliability, safety,
  security and privacy," cautions Peter Neumann, a computer scientist
  at SRI International Inc., a Menlo Park, Calif., think tank. Adds
  Edward Leiblein, formerly the Pentagon's top software expert: 
  "Software problems have reached crisis proportions." The U.S. Defense 
  Department faces a daunting task making its weapons software bug-free,
  partly  because of the sheer size of the job. The Pentagon estimates
  that it will spend $30 billion in 1990 on weapons software, about
  three times the 1985 total. But it figures that demand for software
  is growing so rapidly that by 1990 the industry will be short one
  million programmers and analysts to do the  work. "Any software system
  that is large and has to work perfectly the first time has two
  strikes against it," says John Shore, a former naval software 
  researcher. "But in addition, Star Wars would face a determined
  adversary - that's strike three." Sometimes fixing a small bug can
  lead to greater problems. 

     Two years before the first launch of the Space Shuttle (STS-1), a
  programmer changed the timing on some Shuttle software by one-30th of
  a second. Unknown to the National Aeronautics and Space
  Administration, that minuscule change introduced a 1-in-67 chance
  that the Shuttle's five on-board computers wouldn't work in sync.
  Twenty minutes before the launch in April 1981, the bug appeared, the
  computers couldn't communicate, and NASA scrubbed the flight.  The
  Pentagon has tried to reduce its software problems by requiring 
  contractors to use a single new computer language, called Ada.
  Previously, weapons contractors used as many as 80 different
  languages. But critics such as David Parnas, a weapons-software
  expert at Queen's University in Kingston, Ontario, says that Ada
  frustrates software programmers because it is so difficult to
  understand. Software written in Ada, he adds, runs so slowly that  it
  isn't suitable for weapons, which must respond instantly to threats.
  The Pentagon responds that the speed of Ada programs matches the
  speed of other computer languages. 

	{The Wall Street Journal, 28-Jan-87, p. 1}

 <><><><><><><>   VNS Edition : 1246    Thursday 29-Jan-1987   <><><><><><><>

    Certainly on VAX/VMS the speed of Ada programs matches the speed of
    programs written in other languages.  Other vendors may have
    inferior Ada systems, of course.
        John Sauter

    Comments along the same lines of .5 can be found in "The Sciences"
    magazine (Jan/Feb 1987 issue -- never read the magazine before,
    just picked it up on a whim at a grocery store when I saw the Ada
    article).
    
    It has an article called "Ada's Troubled Debut" and starts off with
    a couple examples of how an Atlas rocket (carrying a nuclear warhead)
    had to be destroyed by the range safety officer because a period
    had been mistakenly substituted for a comma in the missile's flight
    control program.
    
    The article goes on with some other similar stories and starts talking
    about how the DoD is trying to use Ada to cure these problems.
    
    I don't like the size (hence complexity) of Ada and agree 100% with
    a quote by Tony (aka C.A.R.) Hoare given in the article:
    
    	Do not allow this language [Ada] in its present state to be
    	used in applications where reliability is critical, i.e.,
        nuclear power stations, cruise missile, early warning systems,
        anti-ballistic missile defnse systems. The next rocket to go
        astray as a result of a programming language error may not be
        an exploratory space rocket on a harmless trip to Venus: It may
        be a nuclear warhead exploding over one of our cities. An
        unreliable programming language generating unreliable programs
        constitutes a far greater risk to our environment and to our
        society than unsafe cars, toxic pesticides, or accidents at
        nuclear power stations.
    
    -Glenn-

	From NASA Tech Briefs Nov/Dec 1986 special issue.

	            'NASA's ADA Connection'

	
	"NASA first became involed with Ada through the office of
	 Aeronautics and Space Technology, and then through Johnson
	 Space Center, selected in 1983 as one of two federal Beta
	 test sites for DoD's newly-developed compilers.
	 
	 Ideally suited for the life-cycle of large, complex 
	 applications, Ada has been selected for the Space Station.
	 One of Ada's strengths is its support for parallel, fault
	 tolerant program modules especially important for the Space
	 Station. At the Goddard Space Flight Center (GSFC), several
	 projects are being used to access Ada's suitability for
	 NASA use and to gain expertise in the language."

	So, who's right about what Ada can do or can't do with 
	complex applications?

	bob	

    	Well, you don't think NASA's going to admit to any screw-ups,
    do you?
    
    	Larry
    

re .8
                                  -< RE 16.8 >-
Maybe Ada really is OK. Maybe it's just the compilers and RTS that
are loosers. . . As someone surely has said: it's what ya make it!

.10> Maybe Ada really is OK. Maybe it's just the compilers and RTS that
.10> are loosers. . . As someone surely has said: it's what ya make it!

    I think that is exactly the problem: compiler theory (and
    implementation) today can not handle a language so large and so
    full of bells and whistles (like Ada) that a reliable compiler just
    can't be written for it. At least not for critical applications
    like controlling a rocket carrying a nuclear warhead or some such.
    
    
    -Glenn-

    P.S.:  Perhaps this discussion should be moved to some other conference.

Re .7:

>    It has an article called "Ada's Troubled Debut" and starts off with
>    a couple examples of how an Atlas rocket (carrying a nuclear warhead)
>    had to be destroyed by the range safety officer because a period
>    had been mistakenly substituted for a comma in the missile's flight
>    control program.

I swear, more spaceships have been destroyed by Fortran programmers
writing DO10I=1.10 than in all 3 Star Wars movies combined.

So, the US was conducting LIVE missile firings before the above-ground test
ban treaty, AND one of them had to be destroyed in flight by the RSO? In
the old days, when the missiles weren't all that accurate?

I'm skeptical.


The quote by Hoare is from his ACM Turing award lecture on how Ada would
never work.  Ha-ha.  My memory of the lecture is that it is full of
claims about what Ada would never be able to do (e.g. compile efficiently,
run efficiently, etc), but has been doing for a while now.

Re .11:

OK, we'll use Fortran for all the missiles that pass over your house.
				/AHM

    re .12
    
    Live missile firings?? Are you suggesting that the US fired test
    missiles with live warheads? Other than things like Project Argus,
    I didn't think the US had flown any live warheads (Argus involved
    detonating fission warheads in the upper atmosphere, to see what
    would happen).
    
    gary

    Re: the last few.
    
    On May 6, 1962, the USS Ethan Allen fired a Polaris A-2 missile
    carrying a 0.5MT warhead towards an area of the Pacific near Johnston
    Island.  This shot was the only flight of a US missile with a live
    nuclear warhead, and the last live test of a complete US nuclear
    weapons system.

    	Did they detonate the warhead?
    
    	Larry
    

    Yes.  (These were the pre-test ban treaty days.)

Thanks for the specific information, George and Gary.  (I forgot about
ARGUS, and didn't know about the SLBM test).

I hope the Fortran-ICBM-RSO story is now quite dead.
				/AHM/THX