AN ASSESSMENT OF THE UTILITY OF THE ARPA NETWORK OF COMPUTERS FOR THE INTERNATIONAL SECURITY AFFAIRS ANALYST

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP79M00096A000500010013-4
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 36
• Document Creation Date: December 16, 2016
• Document Release Date: September 1, 2004
• Sequence Number: 13
• Publication Date: September 1, 1973
• Content Type: PAPER

ra {. Approved For Release 2004/09/23: CIA-RDP
AN ASSESSMENT OF THE UTILITY
OF THE ARPA NETWORK OF COMPUTERS
FOR THE INTERNATIONAL SECURITY AFFAIRS ANALYST
Working Paper
1 September 1973
Principal Investigator: .
Robert A. Young
Sponsored by
Defense Advanced Research Projects Agency
Contract No.: DAHC15-71-C-0201
ARPA Order No.: 2323
Program Code No.: 2D166
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 .=ABiApPoi V6 diWa[Ts 2004/09/23: CIA-RDP79M00096AO00500010013-4
Page
SUMMARY . . . . . . . . . . . . . . . . . . . . . . 1
1. WHO IS THE ANALYST? .
A. Analyst's Background . . . . . . . ... . . . . 2
B. Analyst's Needs . . . . . . . . . . . . . . 3
C. Current Situation . . . . . . . . . . . . . . . . 5
II. WHAT CAN THE ARPANET DO FOR THE ANALYST? . . 10
A. _ Convenience. . . . . . . . . . . . . . . . 10
B. Fle bility . ... . . . . ... ? 12
C. Reliability . . . . . . . . . . .. . . . . . . 17
D. Ease of Use . . . . . . . . . . . . . . . . . . . 19
E. Data Availability . . . . . . . . . . . . . . . . . . 23
III. WHAT WILL THE ARPANET BE LIKELY TO BE ABLE
TO DO FOR THE ANALYST? . . . . . . . . . . . , 24
A. Convenience. . . . . . . . . . . . . . . . . . . 24
B. Flexibility. . . . . . . . . . . . . . . . . . . 24
-C. Reliability . . . . . . . . .. . . . . . . . . . 25
D. Ease of Learning and Use. . . . . . . . . . . 25
E. Data Availability. . ... . . . . . . . . . . . . . . 26
ii
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 TA B pr? ec- I a j2 MR?I : CIA-RDP79M00096AO00500010013-4
Page
IV. ADDITIONAL INFORMATION. . . . . . . . . . . . . 28
A. ARPANET Technical Information . . . . . . . . 28
B. 'ARPANET Resources . . . . . . . . . . . 29
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 The ARPA Network is one of the most important technologicall;t inno-
vative N rays of using computers since their wide spread introduction in
the 1950's. It permits many different users in different fields to expand
the scope of their computing tasks far beyond what could be done with
any single computer system. The ARPANET has provided practical
assistance to some users with particular research problems. It is now
appropriate, however, to begin finding ways to use the capabilities of
the ARPANET in operational areas of the U. S. Government, particularly
in the Department of Defense, the sponsor (through the Defense Advanced
Research Projects Agency) of the Network.
One class of potential users of the Network has a certain set of problems
which may not be faced by other users or potential users. The community
of international security analysts - individuals who evaluate information
about the behavior and capabilities of foreign nations - has a real need for
the Network; a need which is not currently being filled. This paper at-
tempts to investigate. ways by which that need can be filled.
The report is prepared with the support of ARPA Contract No. DAHC15-
71-C-0201, Dr. Robert A. Young, Principal Investigator; Dr. Michael
R. Leavitt, Project Director. In preparing this report we have had the
gracious assistance of Mr. Jean Iseli of the MITRE Corporation,
Mr. Michael A. Padlipsky of M. I. T., Ms. Barbara Noble of UCLA and
all of those whose reports, notes and information in the Network Infor-
mation Center have proven 'invaluable.
i
Approved For Release 2004/09/23 : CIA-RDP79M00096A000500010013-4
 Approved For Release 2004/09/23: CIA-RDP
5U-%I tiRY
The thesis of this paper is that although the ARPA Network of Com-
puters is a most versatile tool for many different kinds of problem-
solving, it does not yet have the capabilities of serving the analyst in
the international security area. The analyst has a need for convenient,
reliable, relatively easy to learn, flexible systems, and for a moder-
ately wide variety of programs and data. The ARPANET now provides
convenience and flexibility, but not sufficient reliability, ease of
learning, or data.
The paper begins by considering who the analyst is - his background
and technical training, his computational needs, and how he currently
satisfies those needs. The analyst in the international security area.
is a well-trained person whose skills have proven useful in understanding
and recommending action about international affairs. He is trained as
a foreign affairs generalist, but not in the use of computational tools.
Moreover, he is not certain of the need for those tools in his profession.
He works under substantial time pressure with a limited research/ana-
lysis staff and rarely has the time to commence major computational
tasks with new data collections to answer operational questions posed
to him. He requires reliable systems - systems in which he has con-
fidence, and systems which will work -,,,,hen called upon. He needs sys-
tems which he can learn to use himself, and which, therefore, will not
change radically over relatively short periods of time.
The analyst currently has available moderately large batch processing
systems - occasionally he will have access to a time-sharing system..
They ...iil be as reliable and stable as he needs to perform some functions,
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 Approved For Release 2004/09123: CIA-RDP79M00096AO00500010013-4
but they take too long to use effectively. His'. current systems are gen-
erally inconvenient and difficult to learn, and most important, they
rarely have a stockpile of accessible data which can be used to help
answer his questions.
The ARPANET can do much to ease the convenience problem by pro-
viding time-shared access to a very large number of computers. The
flexibility of the Network in permitting the use of a much larger num-
ber of programs, particularly those at UCLA, M. I. T. and Stanford
University, can be a great -aid. UCLA has a very substantial number
of different statistical and modeling programs; M. I. T. has the CASCON
system for local conflict, and the Consistent System for information
processing in the most convenient form; Stanford has an analysis pack-
age linked to the Associated Press wire service.
However, theARPANET is not yet sufficiently reliable for operational
problems. Unreliability can come both from the Network itself, through
its TIP's, IMP's, and channels; as well as from the individual computers
on the Network. Although reliability has improved in both areas over
the last few months, the analyst's need to be as certain as possible that
a system will be available cannot yet be fulfilled. Nor is the Network
particularly easy to learn to use. Connection from a local host compu-
ter obviates most communications problems while connection directly
from a terminal may be difficult. The protocols for common user pro-
grams at various sites exist, but not in sufficient number to permit an
analyst to query the "files" as he would with manual means. If answering
a particular problem requires a major data collection and installation ef-
fort, as it would on most Network computers, the analyst will not use it
to answer those questions, and will rely on his traditional sources.
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
Finally, the Network can be expected to develop its capabilities in all
of these areas over. the near future, thus increasing its utility to the
analyst as well as the probability of its use. The two areas in which
difficulties can be expected to remain are the reliability of local host
computers, particularly those with primarily academic responsibilities,
and the availability of data for the analyst's use. The paper concludes
with a summary of literature about the Network.
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 09/23: CIA-RDP79M00096AO00500010013-4
The purpose of the paper is to assess the ARPANET's utility for the ana-
lyst in the international security community, so it is first necessary to
understand some of the characteristics of the user. The information in
this section is based on observations made during joint work with many
analysts over the past few years as well as on various descriptions of the
analyst and his functions. The section will proceed by first examining
the analyst's background and technical training and will. show that unlike
individuals in other fields (systems analysis, operations research, eco-
nomic policy, military strategy, etc. ), the international:.affairs analyst.
has relatively little technical training - training in the use of analytical
tools :. but an increasing need; for such tools and training in them. What
the analyst needs to perform his job adequately will then be considered
as far as computational support is concerned. In many areas, the ana-
lyst works in a very small group, in a crisis atmosphere and has little
time for detailed research and learning about how particular systems
work. _ His current computing situation will be discussed and its satis-
factory as well as unsatisfactory aspects elaborated.
A. Analyst's Background
The International Security Analyst is a well-trained person whose skills
have proven useful in understanding and recommending action about in-
ternational affairs. These analysts have been properly considered to be
some of the most valuable and competent individuals in the government
for their ability to read, understand,. synthesize and advise. Their role
in government is now changing and, partially because of their past suc-
cesses, analysts have not changed to fit the role.
z
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 Approved For Release 2004/09/23 : CIA-RDP79M00096AO005AOQ10013-4
Specifically, today's analyst is trained as a foreign a pairs generalist
with at least one advanced degree in some area of international affairs,
whether he is a civilian, a military officer or a foreign service officer.
The changes in methods of analyzing international affairs that have talken
place over the last five to fifteen years have, in many cases, been too
recent. for the analyst to make direct use of. He is ambivalent about
their utility for his operational responsibilities and, being well-trained
in the substance of his field, is often hesitant to spend the time required
to learn the newer techniques, which include the use of computers for
statistical analysis, data display, modeling, and sirnulation.
It is most important to note that a major reason that the analyst is un-
willing to spend great amounts of time on these techniques is that they
have not had unqualified success in his field. The techniques are still
controversial within the intellectual community of international affairs
to which he belongs,, and because of time pressures and career possi-
bilities, an excursion into the realm of the controversial is not often
rewarding. His ambivalence is not a matter of unintelligence or ob-
stinacy, but rather, a realistic appraisal of where his valuable intellec-,
tual resources should be spent. His position is thus that proponents of
new ways to make his work more efficient and more productive should
come to him, rather than he to them.
B. Analyst's Needs
The analyst in the international security area is called upon to study a
particular situation and produce recommendations for action by the
U. S. Government. Although not usually in a" crisis atmosphere, the
analyst invariably works under time pressures, with little opportunity
or incentive to "try something new" to see if it works. Further, the
analyst rar^ly has a very large staff worl:ing for him to aid in his work.
Approved For Release 2004/09/23 : CIAJ~RDP79M00096A000500010013-4
 He is expected to have the general store of information he needs in his
Approved For Release 2004/69/23 : CIA-RDP79M00096AO00500010013-4
own mind and his files, and to know where to. go to get whatever else
he needs. The analyst thus needs reliable sources of information -
quickly and surely available - with. an ease of access appropriate to a
small staff. Any innovation that would attempt to serve that analyst will
need to fulfill those particular requirements at a minimum.
The reliability of information sources has two components: the analyst
must have confidence in them, and they must be available when needed.
Confidence requires either that the analyst must have used them before
or that someone must have demonstrated convincingly that they would
have been at least as useful as personal prior use. Currently the ana-
lyst uses reliable sources, in that his colleagues use the same data and
data-acquisition methods. The "files" which contain past actions of his
agency, prior recommendations by him or his predecessors, and the
safety of tradition. are the primary sources of the analyst's information.
To depart from these sources, the analyst needs to be convinced that they
merit both his confidence and the confidence of his superiors.
But no source will be reliable if it is not available when needed. Time
pressures on the order of hours or days prohibit the analyst from relying
on information that can be acquired in a Week, or which is currently
available but he cannot access it for technical reasons. The current
situation permits him or his assistant to go "into the files" for infor-
mation and to have reasonable confidence that it is there. A system
which is unlikely to yield that information will be distrusted and not used.
A computerized information system might not be available to him during
his normal working hours. Access to the system might require that he
contact two or three technical personnel whose responsibilities include
many other agencies or offices. Access to the information' might require
that he or his assistant learn how to interpret written Information that
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 Approved F pr Release d?0o04/0/23 :stA-Rt~ep7~lll$~10~9~$t5~Q0~(In3 Aa
was designedo e rea y ~, zs z bz anrnc:r.
Access to information might require that the analyst be transported to a
site more than an hour's distance from his office. If any of these con-
ditions exist, the information, no matter how believable, is unlikely to
be used for operational purposes.
One additional characteristic of the source of the analyst's information
needs to be mentioned. The procedures for access must be relatively
stable. Once the analyst has made the investment of time to "learn" a
method of information access, he must be able to rely on using that
method for a reasonable period with no substantial new learning necessary.
In other words, he needs an operational information and analysis system
to help in his operational responsibilities.
C. Current Situation
The above sections outlined who the analyst is and what he needs. It is
now necessary to specify what he has available to fulfill those needs in
one specific area: computing resources for data analysis and presentation,
since this is the major area in which the ARPA Network of Computers can
be of assistance. A brief summary of the kinds of computing facilities
generally available will precede an evaluation of how those facilities ful-
fill his need for reliable, stable, timely, and easy to learn information
systems which provide the tools he needs to get his job done.
The most common facility currently available to the analyst is a large-
scale computer, operating in a "batch processing" environment through
a widely used "operating system. " A large-scale computer might be con-
sidered to be a machine the size of an IBM 360/65 or bigger (IBM 370/155,
UNIVAC 1108, Control Data 6000 series or Cyber 70 series, Honeywell
600 or 6000 viould all be acceptable). Some users might have smaller
Approved For Release 2004/09/23 : CIA-ADP79M00096AO00500010013-4
 r~acl pC9P~G~ J oA~~ Q~${ '3n 1ff 2QPlk9@ 1( (dIl~15fl@flffi~~3~A.low; others
may have access to a time-sharing computer which certainly eases some
access problems. The batch-processing system means that the analyst
needs to have a particular piece of analysis prepared for him, submitted
to the computer, retrieved, corrected, resubmitted and finally used.
This process can take from one-half day to one week to be completed for
Many heavily used systems. The fact that the analyst has access to a
widely used operating system means that he will be able to use several
kinds of analysis software - programs already written for general sta-
tistical analysis or report preparation. This means that the analyst need.
not be a programmer, nor does he need a programmer on his staff.
The system fulfills two of his most important needs: stability and reli-
ability. A large, widely used computer system will not change very much
as far as the analyst is concerned. There may certainly be changes
to the operating system, but they will rarely have a noticeable effect on
the user's operation (except to lower or raise the cost of the jobs he does).
Periodically, his analysis programs will be changed, but most often, the
changes will be compatible with his previous procedures, by adding new
procedures or by simplifying old ones. It is thus a stable system, and,
once learned can be used for a period of time without relearning.
The system is reliable, for the most part, for two reasons. First, he is
likely to have confidence in it. Any data that he uses is likely to have
been collected and put on the system by his own office. Any programs
that he uses are likely to have been used before by enough individuals
in similar positions to minimize the risk of error. It is also reliable
because it is going to be available when he needs it, or at least he will
know if it is not able to be of use for a particular problem. A task which
has a deadline of two weeks and will use precollected data may be per-
fectly acceptable for the use of even the most overloaded, the slowest
6
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 hr. ti UipygVjg9 R5~1 ~ 004/~4~e3thC3 i ~0~ '~~~5 01 1 swill not
even be attempted. In other words, the analyst knows the limits of his
tools.
The system will be unsatisfactory, however, in its ability to provide con-
venience, ease cf learning and sufficient data for handling new problem
areas. The convenience aspect holds more for the user of batch systems
than for time-sharing. A user with an immediate problem will not ap-
preciate waiting two to four hours for output. It will be even more un-
satisfactory to the extent that there were errors in the input which re-
quired resubmission. And it will, be totally unsatisfactory if the analysis
is sequential - if each stage of analysis requires the output from a pre-
vious stage. The situation is further degraded if the computer is some
distance from the analyst's office or if the analyst must communicate
his needs through an intermediary, be he an assistant,- an in-shop con-
sultant or worse, an outside consultant.
The time-sharing user has the advantage of much more, rapid return on.
his efforts thus facilitating sequential analysis, ameliorating the affect
of input errors and, usually, shortening the distance to his interface with
the computer. For this user, the current situation is likely to be satis-
factory as far as convenience is concerned.
Both the batch and the time-sharing user have the problem of training.
Although there are an increasing number of exceptions, the effective use
of current computers requires substantial training. One must understand
what a computer can do, what its limitations are and what kinds of ques-
tions one can ask of it just to begin. Many analysts do not have this level
of understanding, without which it is very difficult even to use a com-
petent assistant. To make the best use of an analysis program, one must
have spent some time learning, preferably from a person rather than a
7
Approved For Release 2004/09/23 : CIA-RDP79M00096A000500010013-4
 Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
manual, just what that program can do, what kinds of input it requires,
and how to interpret its output. If this learning takes place during the
fulfillment of a particular task, particularly if there is a. deadline ap-
proaching,. the experience is certain to be frustrating at best.
Most computer systems now in use have little regard for the relatively
computer-naive analyst. Few efforts are made to equip the programs
with flexible input checking to forgive errors. Thus alearning process
must take place. This process is not the defensible one of learning how
to do a particular kind of analysis, but rather how to coerce a particularly
ignorant tool to do that analysis for you. Most analysts have neither the
time, the inclination or the perceived need to undergo that learning.
Finally, most current computers do not have libraries of data comparable
to the "files. " The "files" are those sources of information that the
analyst has relied on for the information on which analysis is based. They
may contain numbers, previous reports, intelligence information, or the
results of prior analysis. To accomplish a new task, an analyst can gen-
erally rely on the files" for the bulk of his data, leaving perhaps a small
amount of information yet to be collected. The "files" are simply not
available on any computers in the international security community (with
some exceptions), so the analyst must place any needed information on
the machine himself. After a period of time, the analyst may build up
his own computerized "files, " but their initial lack provides a barrier
to current use. The process of building reliable information is difficult
in any situation; on computers, however, the difficulty is compounded by
the need for precision at every stage of the process.
As a result of the current computing situation, then, international security
analysts are partially satisfied but generally not encouraged to use some
of the newer tools in their operational problems. The ARPA Network
Approved For Release 2004/09/23 CIA- RDP79M00096AO00500010013-4
 Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
offers much potential for resolving these problems leading to greater use
of computer related analysis methods. The following section details how
the ARPANET may be useful and where it too is unsatisfactory.
Approved For Release 2004/09/23 : C1A9RDP79M00096A000500010013-4
 II. I' K "sA) .qqA/R :S*fM7WgQ W OOq 0013-4
The thesis of this section is that the convenience and flexibility of the
ARPA Network can be of some use to the international security analyst
today, but that it is not yet reliable enough, sufficiently easy to learn and
to use nor well enough stocked with data to permit its adoption'. Although
the emphasis of this section is negative on balance, the final section will.
illustrate that the current Network situation is still changing, and that
the expectation-is that the future of the Network is real potential for the
analyst.
A. Convenience
The major purpose of the ARPA Network of computers is convenience and
flexibility:
The goal of the computer network is for each computer
to make every local resource available to any computer
in the net in. such a. way that any program. available to
local users can be used remotely without degradation. That
is, any program should be able to call on the resources
of other computers much as it would call a subroutine.
The resources which can be shared t its way include soft-
ware and data, as well as hardware.
On the whole, the Network has succeeded in its goal. It is convenient in
two distinct ways. If a user has local access to a "host computer, " he
'Larry G. Roberts and Barry Wessler, "Computer Network Development
to Achieve Resource Sharing, " in AFIPS Conference Proceedings, Vol. 36,
1.970 Spring Joint Computer Conference (Montvale, N. J.: AFIPS Press,
19-70), p. 543.
10
Approved For Release 2004/09/23 : CIA-RDP79M00096A000500010013-4
 can ApavedtFar Rekea-se Q??41E1 IZi:aIAAKMO ltl?0896A@ dt0-.eve, The
access is achieved by a network of high-speed communication channels
which links over twenty-five different computers throughout the United
States, Hawaii, and most recently England and Norway. By an appropriate
use of the local computer's "file transfer protocol" program, a user can
transfer large amounts of data from any other computer on the Network
to his own for processing. Thus, if an analyst at the RAND corporation
wished to perform a statistical analysis at his own site using data normally
available only on the MIT-Multics computer, it would be a relatively
simple matter to have the data brought over specifically for use by his
The more striking convenience comes for the user who does not have direc
access to his own computer, however. This user may use a standard
telephone line to call a Terminal Interface Message Processor (TIP) whits
can then connect to a computer on the Network. Some TIP's are for the
sole use of their owners (for example,. TIP's at the USN Fleet Numerical
Weather Central in Monterey, or at the USAF Environmental Technical Al
plication Center at the Navy Yard Annex in Washington). Others function
(currently) as a more general resource (for example, TIP's at MITRE in
McLean, Virginia and at the National Bureau of Standards. in Gaithersburg,
Maryland). The analyst with access to a TIP from a communications ter-
minal (such as a Teletype, an IBM 2741, a General Electric Terminet or
a Hazeltine 2000) can "look like" a local time-sharing user of any "server"
computer on the Network. As must-other local users, he must make ar-
rangements for access with the local computer personnel, but this is
facilitated by the presence of a Network Liaison person at each site. 'In
addition, there are several servers which provide experimental usage of
their systems at no dirert charge to the user (these include MITI -multics,
Stanford's Artificial Intelligence Laboratory, and the Network Information
Center at Stanford Research Institute).
11
Approved For Release 2004/09/23 : CIA-RDP79M00096A000500010013-4
 1 th;:Ap Dp& FFbrl ~q~~Cs @O IO~~ a 6 f)P7J b0!1bA 60 1b0il-,pPerational
taslk are on the Network, the convenience of using that information, either
at the location of the information, or at the analyst's home site (if a mem-
ber of the Networl=.) cannot be surpassed.
B. Flexibility.
Flexibility in this context means the ability to perform or to help in the
performance of a wide variety of computational tasks. The Network gives
the analyst more different computer programs and ways of using those
programs than he will be able to use. (The Network is so flexible in this
sense, that it almost becomes a disadvantage, as discussed in the next
section. )
Of the over twenty-five computers currently on the Network designated as
"servers, " a few provide the bulk of the readily available programs that
an analyst is-likely to need. These are the Campus Computing Network
at the University of California, Los Angeles, the Artificial Intelligence
Laboratory at Stanford University and the Multics computer, a part of
the Information Processing Center at the Massachusetts Institute of
2
Technology.
1. Campus Computing Network at UCLA
The UCLA Campus Computing Network (UCLA-CCN) is served by the
largest and currently most powerful computer on the Network, an IBM
360 model 91. The size of the computer permits readily available
2MM1any other computers have many other resources of general interest;
however, for the current and projected future needs of international
security affairs analysts, these three are of the greatest interest.
12
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 Approved For Release 2004/09/23 : CIA-RDP79M00096A000500010013-4
storage of some of the most commonly used statistical and modeling
programs available. Its statistical packages include the set of Biomedical
Computer Programs (BMD) ,hich, in spite of its name, is perhaps the
most commonly used set of programs for nearly all kinds of statistical
data analysis. BMD is not the easiest program set to use or interpret,
but it is likely that if an analyst is familiar tivith only one set of programs,
it will be BMD. UCLA-CCN also has the Statistical Package for the
Social Sciences (SPSS) available. SPSS is substantially easier to learn
and to interpret because it is a single integrated package. DATA-TEXT,
(a package developed at Harvard) and the Statistical Subroutine Package
(SSP, an IBM product) are also available..
UCLA-CCN also has the two most common computer simulation programs
available. SIMSCRIPT II. 5 (a proprietary simulation language developed
by C. A. C. I.) and GPSS (an IBM product) can be used by the experienced
user for a very wide variety of tasks.
UCLA-CCN permits two modes of operation: time-sharing and remote
batch processing. Both modes can be done from a computer terminal,
although they have very different characteristics. Time-sharing is per-
formed under the Time-Sharing Option (TSO), with which the-user may
write programs in FORTRAN, PL/1, and BASIC. The time-sharing user
may create and edit files of data or textual information, and use these as
input and to store output from programs. Remote-batch processing is
done from the terminal, but in this mode of operation, the user creates
a file of information which the computer processes independently of user
interaction. It is as if the user created a batch processing job and sub-
mitted it to the computer. As with batch jobs generally, the user must
wait for the computer to finish processing before he is able to examine
the output. The system is faster than normal batch processing, but the
lacy. of dirc:ct user intervention can be a problem. Most of UCLA-CCN's
Approved For Release 2004/09/23 : CIA RDP79M00096A000500010013-4
 ? p r og r Approargd fm)teReleetse: 280 12312 UP7.-O 1AdioU61kt00?SOUDR
beginning of this subsection.
jgr4ed at the
2. Stanford University - Artificial Intelligence Laboratory
At first glance, there is no particular reason why an artificial intelligence
operation should be of use to an international security affairs analyst.
However, one of their projects involves making the Associated Press
Wire Service available to users of their system. The AP wire is avail-
able in two ways. Through a program named "HOT, " the user may have
his terminal behave like an AP terminal, with national and international.,
news stories printed out at the same time that they appear on terminals
in news rooms across the country. This is more a form of entertainment
than a useful adjunct to analysis. The useful product is a program en-
titled "APE" (Associated Press Extractor) which retrieves information
from a constantly updated one-day file of AP stories. Information can be
retrieved by topic ("SALT TALKS, " "MIRY, " "ABM"), by country or city
("USSR, " "MOSCOW, " "WEST BERLIN") by individual ("NIXON, " "MAO, "
"SADAT") or by any combination of words that may have appeared in a news
story. Once retrieved, the information can be displayed at the terminal
or stored in a file. Since only one day's worth of stories is available at
any time, it is necessary to perform the analysis daily, or take advantage
of the "standing order" facility by which stories. of interest are automa-
tically stored in a user's file. Files thus created can be read from the
computer, or sent, via the File Transfer Protocol program, to another
computer for content analysis or long-term storage.
Stanford Artificial Intelligence Laboratory (SAIL) is one of the systems
which provides free experimental usage, thus facilitating investigation prior
to commitment of resources. Since it is located in the Pacific Time Gone,
it is relatively unused during the morning (an the East Coast) and thus an
14
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 attraA rN9 es s qe ~tlea 0 /Q 4 1' cJMQ.ETY4R%OW 0,_, 992 ~.0Ds 3:4~1z-t~. ritt e
and brief. 3
3. Multics at the Massachusetts Institute of Technology
MIT has four different computers currently on the Network, Multics being
the most often used as well as the most powerful. Multics began as an
experimental system to investigate the idea of a "computer utility" a
system which. could simultaneously support, at different levels of service
and different pricing arrangements, a large number of users. Multics
has recently been moved from an experimental computer (a Honeywell
Information Systems 645) to a production version which is being com-
mercially marketed (HIS 6180). It is of great interest for its means of
protecting the files and security of its users and its effective "unlimited".
storage capacity (from the user's perspective), but for the international
security analyst, two particular programs are of interest: CASCON and
the Consistent System.
CASCON is a single program which provides systematic information about
52 local conflicts since the second World War, to the user at the terminal.
The user can enter information about new conflicts to determine how com-
parable they are to the older ones (the "file"). Basically an information
retrieval system, CASCON4 was developed for the U. S. Arms Control
and Disarmament Agency, and has not been used to any great extent
3See Martin Frost, "Reading the Associated Press News" (Stanford,
California: Stanford Artificial Intelligence Laboratory, July 1973).
4Developed by Lincoln Bloomfield and his associates at MIT's Center
for International Studies. See Lincoln Bloomfield, Robert R. Beattie,
and G. Allen Moulton, "CASCON II, Computer Aided-System for Handling
Information on Local Conflicts: User's Manual" (Cambridge, Mass.
Center for International Studies, September 197Z).
15
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 becaq f'eAtRoIaR@& -2904Mti34l@IA-R4P N$6(ON 069l"Obl g-operation
(although it has been converted for use on the" University of Michigan
system, not a member of the ARPANET). CASCON may prove of in-
terest to some members of the international security community.
The Consistent System, on the other hand, is a large set of programs
designed to help behavioral scientists use computers in their research.
A part of the ARPA-funded Cambridge Project, the Consistent System
is available only on Multics. Of particular interest for data analysis and
presentation is the JANUS subsystem of the Consistent System. JANUS
permits a limited number of kinds of statistical analysis, but is one of
the most flexible, forgiving programs currently available. Other functions
that the Consistent System is able to fulfill include programs for computer
content analysis of documentary material and, for the more sophisticated
user, a very complete system for classifying and manipulating stored
information.
Many other kinds of programs are available on the Network, but those
l4sted above may properly be considered as the most relevant to the
needs of most international security analysts. It is fair to say that they
represent a larger base of computing power than is currently available
on the large operational computers in the various agencies.
One important inflexibility remains to be noted that will prevent the
ARPANET from replacing currently used computing resources for some
tasks. The ARPANET is, of course, an insecure network. No classified
analysis can be done directly through the Network nor can,,it be -done on
the computers in the Network, unless the analysis is masked in such a V ay
.Y.
as to-provezt'anypossibility of determining the content of the analysis.
For these purposes, analysts will need to continue to use secure instal-
lations in the manner prescribed by the appropriate regulations.
16
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 C. R4Rpgq)NI or Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
The ARPANET is not yet reliable enough to serve as an operational tool.
It is unreliable both in its communications system, and in its components,
the host computers. This strong statement admits to some exceptions,
but is based on the rather unusual needs of the international security ana-
lyst. As elaborated above, an analyst can be called upon for a report
which is due a matter of days or hours from the time of its request. For
any computer system to be of use in this environment, it must be able
to guarantee its own reliability when called upon. The ARPANET cannot
yet do that for two reasons: unreliability of the Network communications
and unreliability of the hosts.
1. Network Unreliability
The Network is a complex linking of computers through intermediate com-
puters through which data are transmitted.. The TIP discussed above is
a generalized version of the Interface Message Processor (IMP) a com-
puter which receives information from the various host computers to be
passed on to other IMPTs and then to their proper destination. Given the
number of stages through which a particular set of data must pass to reach
its destination, the ARPANET is reliable for most purposes.5 However,
the Network can become unavailable to the user for a variety of reasons,
including, malfunction of his own terminal, his host computer, or the IMP
at his host. The first two problems cannot properly be attributed to the
5 See Howard Frank, Robert E. Kahn, and Leonard Kleinrock, "Computer
Communication Network Design - Experience with Theory and Practice, "
in AFIPS Conference Proceedings, Vol. 40, 1972 Spring Joint Computer
Conference (Montvale, N. J.: AFIPS Press, 1972), pp. 255-270 for a
more complete discussion of the reliability problems in the ARPANET.
17
Approved For Release 2004/09/23 : CIA-RDP79M00096AO00500010013-4
 -
t
i\et