arms-d@ucbvax.ARPA (06/26/85)
From: The Arms-D Moderator (Harold Ancell) <ARMS-D@MIT-MC.ARPA>
Arms-Discussion Digest Volume 3 : Issue 51
Today's Topics:
Provisions for Detecting Nuclear Bombs
SDI Software: Quanity of Code, Fletcher Report, Abstract of
Paper, and What about a spacewar autopilot?
Nuclear Terrorism & Doomsday Machines
----------------------------------------------------------------------
Date: 25 Jun 85 08:07:02 EDT
From: DIETZ@RUTGERS.ARPA
Subject: Can we detect nukes?
Yes, I've heard the US government has a group stationed in Las Vegas
that responds to nuclear threates, and that they have sensitive
radiation detection devices (both airborne and truckborne) for
locating nukes. Details are of course not available, but I would
think a sensitive gamma ray spectrometer could detect nuclear weapons
at a considerable distance. Plutonium based weapons should be more
detectable than uranium-235 based devices, since they are contaminated
with Pu-240 and 241 which decay with relatively short half-lives. I
think the false positive rate is rather high, though, since there are
lots of weak sources of gamma radiation out there, and a determined
adversary could add more by spreading some low-level material around.
------------------------------
Date: Tue, 25 Jun 85 09:30:08 pdt
From: Mark Stout <stout@Lancelot>
Subject: Detecting Nuclear Bombs
There exists a team of specialists under the control of the federal
government called NEST (Nuclear Emergency Search Team) whose job is to
deal with nuclear terrorism. They deal with both threats to nuke
places and just to spread radioactive contamination. According to an
article I read about them a while back they do have devices capable of
detecting the presence of nuclear devices.
I hear also that such devices can be carried on SR-71s.
I do not know anything about how these devices might work; physics
isn't my long suit.
--Mark Stout
------------------------------
Date: Tue, 25 Jun 85 09:46:17 EDT
From: Herb Lin <LIN@MIT-MC.ARPA>
Subject: Quanity of Code
To: HGA@MIT-MC.ARPA
Does anyone have any idea what DARPA is planning on doing with 10
million lines of code? That strikes me as a completely excessive
quanity; perhaps some dreamers really are planning on taking men
completely out of the loop, something I really doubt we'll ever
do.
No one is talking about taking men out at the strategic level where it
will be turned on. They all say that men are out at the tactical
level, when you have 30 min to process 50,000 objects. FYI, the
Safeguard ABM system required 3 M lines, and the software for NORAD is
about 3 M lines too.
One thing to remember when considering the undoubtly large quanity
of code to be written: while it is clear that the battle
management code can't be given a real test (which is a strong
argument for having men in the loop to correct problems in real
time) the code for specific weapons can be fully tested ahead of
time.
No time to correct the battle management code in real time. The
estimates I have seen say that weapon control software (vs BM) is only
about 10% of the entire system.
[From the Moderator: I was actually suggesting men in the loop to
correct resource allocation problems, like "OMG, nobody's covering the
Foo missle fields...." I'll respond in more detail after I read your
paper.
- Harold
]
------------------------------
Date: 25 Jun 1985 11:42-EDT
From: Hank.Walker@CMU-CS-UNH.ARPA
Subject: Re: Arms-Discussion Digest V3 #50
The only unclassified part of the Fletcher report is the chapter on
battle management. This basic structure of the system would include a
distributed database of targets, status of the system, status of
communication links, target discrimination, target assignment, etc. A
friend who has read this chapter thinks that the 10M line estimate is a
reasonable one. It was based on the basic system architecture and
other large software systems like the shuttle, WWMCCS, and Safeguard.
The problem is not testing the code for any specific component. We can
probably have fairly high confidence that individual pieces such as
low-level signal processing code works. The problem is indeed that the
system will probably have to have a reasonable amount of coupling.
Each individual battle station will have to have a set of
fault-tolerant computers just like the shuttle, so it is indeed subject
to the same sorts of errors. I would think that the real problem is
writing software that will have the desired behavior under all sorts of
not-well-understood situations such as nearby nuclear explosions, parts
of the system going arbitrarily flaky, etc. The distributed database
will be tough since it must be globally distributed and updated very
rapidly.
------------------------------
Date: Tue, 25 Jun 85 19:27:45 EDT
From: Herb Lin <LIN@MIT-MC.ARPA>
Subject: SDI Software...
Here is an abstract of a paper recently completed. It is a revised
version of a paper circulated earlier this year. If you want a copy,
pls let me know. ARPANET people can get it by FTP. Others should
probably send Snail Mail for a hard copy, because it is too big for
most mailers (there will probably be an invoice associated with it for
the cost of reproducing an 80 page document).
Herb Lin
Center for International Studies
E38-616
M.I.T.
Cambridge, MA 02139
(617) 253-8076
Software for Ballistic Missile Defense
June 1985
Abstract
A battle management system for comprehensive ballistic missile
defense must perform with near perfection and extraordinarily
reliability. It will be complex to an unprecedented degree,
untestable in a realistic environment, and provide minimal time
for human intervention. The feasibility of designing and
developing such a system (requiring upwards of ten million
lines of code) is examined in light of the scale of the
project, the difficulty of testing the system in order to
remove errors, the management effort required, and the
interaction of hardware and software difficulties. The
conclusion is that software considerations alone would make the
feasibility of a "fully reliable" comprehensive defense against
ballistic missiles questionable.
IMPORTANT NOTE: this version supersedes a widely circulated but
earlier draft entitled "Military Software and BMD: An Insoluble
Problem?" dated February 1985.
------------------------------
Date: 25 Jun 1985 0509-PDT
From: Rem@IMSSS
Subject: SDI? What about spacewar autopilot?
It seems to me Reagan's strategic-defense plan is orders of magnitude
more difficult than writing an autopilot for the classic video game
"spacewar". Yet as far as I know nobody has written the spacewar
autopilot. Perhaps a video-game autopilot that can shoot down
hundreds of incoming weapons/vehicles would be a good starting point
to discover if such kinds of software is within the state of the art.
Does anybody know about such efforts, either successful or
unsuccessful?
------------------------------
Date: 24 Jun 1985 19:24 EDT (Mon)
From: Wayne McGuire <MDC.WAYNE%MIT-OZ@MIT-MC.ARPA>
Subject: Nuclear Terrorism & Doomsday Machines
From: Herb Lin <LIN@MIT-MC.ARPA>
...the problem is getting the nuclear explosive, not
transporting it. Terrorists can get the bomb either by buying it,
stealing it, or making it. We can't stop any nuclear power from
selling one (or giving one) to someone, but I think the nuclear
powers are responsible enough to refrain from this one.
And yet how much responsibility has the U.S. shown in allowing Israel
to acquire the bomb by siphoning off, through theft and fraudulent
tactics, American nuclear technology and materials? (An occasional
slap on the wrist does not count as responsible restraint.) If the
Soviet Union now looks the other way while one of its client states
pilfers its nuclear weapons technology, how credibly can we protest?
The purpose of Permissive Action Links is to keep a stolen
bomb from going off (it gets disabled when the wrong code
is punched in). Finally, bombs are damned hard to make, even if
the raw material is (relatively) easy to get.
There is a great difference between being /impossible/ to make and
being /hard/ to make. Because the incentives for possessing these
devices are so great, if they are theoretically buildable by small
nations and groups, then eventually they will be built. And with each
passing year, the knowledge and means for constructing nuclear devices
inevitably seeps downward and outward from the tight control of a
select scientific priesthood to more and more random people at large.
With regard to the question of sabotage and theft, the following
paragraphs also appeared in the _Times_ article:
[begin]
Rear Adm. Thomas Davies of the Navy, retired, said, "When we look
at the terrorist record of the past few years, it is safe to conclude
that sabotage of military and civilian nuclear facilities is at the
top of the danger list."
"The spectrum of targets for sabotage--at mines, enrichment and
reprocessing plants, reactors, storage facilities, waste sites--is
very broad, and the consequences of destruction or damage range from
unpleasant to cataclysmic."
The admiral, formerly in charge of nonproliferation matters at
the Arms Control and Disarmament Agency, contends that the
possibilities of sabotage directed at commercial nuclear plants is
"infinite."
He adds, "The risk of theft of special nuclear materials and of
weapons or their components is also very real." Explosive materials
are "always traveling--moving by air, sea, truck, and railway from the
mines to the enrichment plants, the fabricators, bomb assembly depots,
power reactors, processing plants, and storage."
"Transport of so much dangerous material in open commerce may
well turn out to be the Achilles heel of the nuclear industry," he
says, "a number one target for terrorist theft."
[end]
Mr. O'Keefe, author of the book ``Nuclear Hostages,'' an
examination of nuclear war and nuclear terror, says, ``I
believe that the greatest threat to civilization today is
the prospect of a terrorist-implemented nuclear explosion.''
In my view, this is just nonsense. I believe that this is the
most likely possibility for a nuclear bomb to be used, but to believe
that it would mark the end of civilization is absurd. It does NOT
threaten civlization in the way that a 10 gigaton war does.
Detonating ten or twenty devices might not end civilization, but it
would mangle it beyond recognition.
(By the war, Nuclear Hostages is not particularly compelling.
O'Keefe makes several wrong statements in it, and is not very
convincing about politics, strategy, or military affairs.)
I must confess that I have not yet read Mr. O'Keefe's book, but I do
know that he has enormous credibility. He is very highly placed in
the defense/intelligence communities, and probably knows as much about
nuclear weapons as anyone in the world. Until I've been able to
compare his detailed arguments against yours, I'm inclined to treat
his remarks with respect.
A small nuclear crazy state could not strike at the entire world
-- only the US and the SU can. They might strike at someone, but
not everyone.
Mr. O'Keefe, and some other experts, seem to be asserting that small
nuclear crazy states or groups /could/ hit, or credibly threaten to
hit, many cities or strategic sites around the world, certainly enough
to create an unimaginable disaster. That possibility is the point of
the article.
Perhaps we should be worrying a good deal more about the
potential behavior of small crazy states and groups, especially those
motivated by religious extremism and apocalyptic belief
systems, and somewhat less about the plans and actions of the
superpowers.
One nuclear detonation will change the world, but it won't destroy
all of us. I am worried about nuclear terrorism, and even believe in
pre-emptive strike to eliminate it, but I repeat, the main problem
is 10 gigatons and no way of crontolling them.
We are not talking about one detonation, but many. If a small nuclear
crazy state can smuggle one device into, say, the United States, it
could just as easily smuggle in ten or more devices. How about two
devices each in Manhattan, Washington, D.C., Boston-Cambridge,
Chicago, and Los Angeles: exploding these devices would not destroy
the world, but it would bring the U.S. to its knees. The same tactic
could be used against the Soviet Union.
I think the least likely scenario for the start of World War III is
that the U.S. or USSR will launch a nuclear attack on the other; both
sides have too much to lose, and are (with some notable lapses)
fundamentally rational. The most likely scenario is that a small
state or group, motivated by religious messianism or some other
irrational impulse, and with very little or nothing to lose, will
massively attack one or more of the superpowers. Even a "small"
attack--with just a handful of devices, and perhaps against a small
neighbor--could set in motion a larger conflagration.
Analysts who are concentrating exclusively on scenarios of
conventional warfare or nuclear conflict between the U.S. and USSR,
and who are diddling with the comparative minutiae of U.S./Soviet
strategic inventories, are, I think, misdirecting their attention.
Their eye is not on the ball. There may be an analogy here to the
failure of some makers of computer mainframes to underestimate the
impact of minicomputers and microcomputers. Briefcase bombs in the
hands of small fanatical powers may be of much greater future
strategic consequence than massive weapons systems controlled by the
superpowers that are for nearly all purposes unusable.
Wayne McGuire <mdc.wayne%mit-oz@mit-mc>
------------------------------
[End of ARMS-D Digest]