1330
PROCEEDINGS OF THE IEEE, VOL. 66, NO. 11, NOVEMBER 1978
Applications of Information Networks
J. C. R. LICKLIDER MEMBER, mEE, ND ALBERT VEZZA, MEMBER, IEEE
In vited Paper
Ab#ract-The present and' projected appUcations of computer-
communication networks ol information networks include eleclonic
mail, teleconferencing, "the office of the futures" management informa-
tion systems, modeling, "computerized commerce," monitoring of
patients, military command and control, home security, education, and
news. This paper briefly examines 30 such applications and the net-
work capabiliiies they require. It presents a way of estimating the
relative importalice of ¾l'ious network characteristics and of predicting
the suitability of a network or netwolk achitecture for a given set of
applications. The paper then considers several issues that relate to the
political, social, and economic impacts of networks. Among the issues
ae privacy, security, compatibility, impact on productivity, the roles of
networks in international technology transfer and econom competi-
tion, and the confluence or collision of the fielets of co >n>Vputers and
telecommunications, /
I. INTRODUCTION
HE SUBJECT of this paper is applications bf networkso
The networks involve the use of computers, but com-
putation in the narrow sense does not necessarily domi-
nate the applications. The scope of the paper includes, no less
than computation, computer-based applications in which the
main emphasis is on communication among people, on access
to information, or on control of systems, organizations, or-to
mention early one of the deepest though least imminent
concerns-societies. The applications of networks that we shall
examine include electronic message communication [1]-[4],
electronic funds transfer [ 5], access to information, computer-
based office work and "telework," management of organiza-
tions and command and control of operations, education,
entertainment and recreation, reservations and ticketing, and
several others.
Some of the problems and issues in network applications are
mainly tcclmical and some are mainly nontechnical,but almost
all arc mixtures o.t the two, and in most of them the tcclmical
and nonteclmical factors interact strongly. For cxa.mple, the
relative merits of circuit switching and packet switching are
mainly a technical matter, but the fact that the electronic
switching stations of the e:dsting telecommurdcations "plant"
arc cLrcuit switches surely is an economic factor in the circuit-
switching/packet-switching issue. The determination of what
should be the individual citizen's right to informational privacy
is mainly a nontechnical matter, but the pragmatics of pro-
viding informational security, the tectmical basis for assurance
of privacy, must enter the decision process. The national
telecommunications plant-in-being of the U.S. figures strongly
in many of these problems and issues and forces them to in-
v61ve both tcch_rfical and nontechnical factors. The plant is
valued at somcthZng like $120 billion, and most of it was de-
sigucd to carry analog voice signals, which are quite different
Muscrip½ received Mach 10, 19'78; re'/,sect July '7, 19'78.
The authors are with the Laboratory for Computer Science, Ma. ssa-
chusetts Institute of Technology, Cambridge, MA 02139.
in their spectral and temporal parameters and in their require-
ments for error handling and security, from digital computer
signals of the kinds that will flow through the networks of the
future. Because of its inherent redundancy, speech remains
intelligible even when mixed with considerable mounts of
noise, but even a single undetected error-a single bit-can
have extremely serious consequences in electronic funds trans-
fer (EFT) or seriously degrade the performance of a network
carrying enciphered information.
One of the major motivations for networking is the need to
share resources. The main resources that are often advantageous
to share are communications faciJ. ities, computer faciJ. ities, and
information itself. The design of a network can make it easier
or more difficult to share resources and thus directly influence
the amount of resource sharing that will occur. The amount
of communication facilities sharing depends upon many design
factors, all of which influence how well the network is able to
allocate resources dynam/cally in response to changing needs
and availabilities. Though the need for sharing certain types
of computational fac/lities may diminish with the arrival of
the age of the personal computer, it is not at all likely that the
need to share resources will disappear altogether. Geographi-
cally distributed users can share, through a computer network,
the costly high-performance computers that are required to
solve certain large computational problems. Even those using
personal computers to satisfy the bulk of their computing
needs may wish to avail themselves through a network of spe-
cial software services provided by vendors-and they will cer-
tainly wish to communicate with one another.
The sharing of information is the most important type of
resource sharing. The term "information sharing" immediately
conjures up the thought of sharing large data banks of in-
formation among many users, but that is only one aspect of
information sharing. All the applications discussed in this
paper have aspects of information sharing. Applications con-
cerned with communications, management, commerce, govern-
ment, protection, education, and awareness all involve sharing.
The convenience and effectiveness with which sharing can be
accomplished and the facility with which information can flow
across the boundaries of individual application programs will
kave profound effects on kow well the applications serve their
intended purposes.
Many problems and issues arise from the interaction of in-
formation sharing with information security. For example,
should EFT have a network or networks of its own to simplify
the problem of providing secure transmission, processing, and
storage of funds data, or should EFT messages be carried over
a general-purpose network so that a reservations and ticketing
operation can be completed in a single transaction involving
the traveler's organization, the airline and the bank.
0018-9219/78/1100-1330500.75 ¸ 1978 IEEE
------------------------------<page break>-----------------------------
ICKLIDER AND VEZZA: APPLICATIONS OF INFORMATION NETWORKS
II. APPLICATIONS
;'In the context of/nformation networks, just as in the con-
't of computer systems, an application is essentially the ira-
of a purpose. Applications, like purposes, may
be defined narrowly or broadly. When the airlines began to
develop computer-based reservations systems and formed a
consortium, ARINC, to interconnect several of theix systems,
the application was narrow: akLine 'reservations. Now, after
yea of growth and augmentation, one can rent a car, reserve
a hotel room, and arrange to be greeted with flowers and
mariachi mudc. The broadened application might be defined
as reservations and ticketing for almost anything that flies to
or can be purchased at a distant place. One can project the
broader definition into the future and envision a general reser-
vations and ticketing application, operating in a nationwide
or worldwide common-carrier network, through which anyone
could examine the availability of, and reserve or buy a ticket
to, almost anything in the broad class to' which reservations
and tickets apply. But, of course, there ks no reason to stop at
that particular class. One can expand the scope further and
arrive at "computerized commerce," dealing with the wltole
garaut of things that carl be bought and sold. The very brqdly
deemed application would include advertising, dynamic pcing,
and computer-based purch_elng strategies. It might even make
a place for cartels of suppliers and cooperatives of consumers.
No doubt there would be vigorous competition among several
or many offerers of the application, and perhaps one can
imagine even a "meta-market," an over-arching system that
interconnects and integrates the competing "computerized
commerce applicatibns."
In any event, that introduces Ihe notion of applications of
information networks. It might serve to introduce, also, the
nfion of issues, which involve the interplay of the opportunity
and the threat aspects of applications. It is not difficult to
imagine the mischief that could be played by pranksters or
dissidents in a poorly protected,p. ublicly accessible, nation-
w/de reservations system.
A. Bae Applications
Computer-communication networks perform three basic
classes of operations upon information: transmission, process-
ing, and storage. The earliest recoied applications of net-
works were essentially separate exploitations of the three basic
clases of operations. Transmission of information through a
network from a program in one computer to a program in
another, of course, requires some processing and some storage
(memory), but in simple message communication and
transfer interest is focused sharply on transmission. In every
practical computer, procesdng requires storage (memory or
registers), but in early time-sharing services such as Quiktran
[6]-[8], which when introduced did not provide intersess/on
f'fle storage, the (dial telephone) network application was
essentially access to processing. In the Datacomputer [9]
service available through the ARPANET [10]-[12], although
processing is involved in both storage and retrieval, one of the
main applications is essentially access to storage in and of it-
self: the Datacomputer is a place to park bits.
A fourth essential network function combines the bas/c
transmission, processing, and storage operations to provide
access to.information-with the focus of interest on the in-
formation 'itself, rather than on any of the three basic ele-
mentar operations.
1331
Simple message communication and fire transfer, access to
time-shared processing, access to storage, and access to in-
formation are important as well as fundamental network func~
tions, but they are no longer typical of the activities or services
we associate with the term "application." In present-day
parlance, "application" suggests something more highly differ-
entiated and specialized and closer to some specific task or
miss/on.
B, Communication Applications
In the dev/!lopmental hitory of the ARPANET, electronic
message service was a sleeper. Even before the network in-
cluded a dozen computers, seve. ral message programs were
written as natural extensl6ns of the "mail" systems that had
arisen in individual time-sharing systems in the early 1960's.
By the Fall of 1973, the great effectiveness and convenience of
such fast, informal message services as SNDMSG [13] had
been discovered by almost everyone who had worked on the
development of the ARPANET-and especially by the then
Director of ARPA, S. J. Lukasik, who soon had most of his
office directors and program managers communicating with
him and with their colleagues and theix contractors via the net-
work. Thereafter, both the number of (intercommunicating)
electronic mail systems and the number of users of them on
the ARPANET increased rapidly.
l J Electronic Mail, Electronic Message Systems: It soon be-
came obvious that the ARPANET was becoming a human-
communication medium with very important advantages over
normal U.S. mail and over telephone calls. One of the ad-
vantages of the message systems over letter mail was that, in
an ARPANET message, one could wtite tersely and type im-
perfectly, even to an older person in a superior position and
even to a person one did not know very well, and the recipient
took no offense. The formality and perfection that most
people expect in a typed letter did not become associated with
network messages, probably because the network was so much
faster, so much more like the telephone. Indeed, tolerance for
informality and imperfect typing was even more evident when
two users of the ARPANET linked theix consoles together and
typed back and forth to each other in an alphanumeric con-
versation. Among the advantages of the network message
services over the telephone were the fact that one could' pro-
ceed immediately to the point without having to engage in
small talk first, that the message services produced a pre-
servable record, and'that the sender and receiver did not have
to be .available at the same time. A typical electronic mail
system now provides a rudimentary editor to facilitate prepa-
ration of messages, a multiple-addressee feature to make it
easy to send the same message to several people, a Erie-inclusion
scheme to incorporate already prepared text files into a mes-
sage, an alerting mechanism to tell the user that he has new
marl in his mailbox, facilities for reading received messages,
and a "help" subsystem. The prospects of electronic mail ap-
pear to have caught the attention of computer manufacturers
and software and time-sharing firms as well as telephone corn-
parties, national telecommunication authorities, and the U.S.
Post Office-and most of them now seem to be planning, de-
veloping, or even offering some kind of electronic mail service.
Even before electronic mail was well established, .it had be-
come apparent that users would need computer aids for
scanning, indexing, f'fling, retrieving, summarizing, and re-
sponding to messages. Indeed messages are usually not isolated
documents but documents prepared and transmitted in the
------------------------------<page break>-----------------------------
1332
course of performing complex activities often called "tasks."
Within task contexts, messages are related to other messages
and to documents of other kinds, such s forms and reports.
It .eem. likely that we shall see a progressive escalation of the
functionality and compreheusiveness of computer systems that
deal with messages. If "electronic mail" refers to an early
stage in the progression, "electronic message system" is appro-
priate for a later stage and "computer-based office system" or
some comparable term for the stage of  integration. At
some intermediate point message service wfil no doubt be
blended with d/rect user-to-user linking to provide for delay-
free conversation whenever both sender and receiver are on-line
at.the same time an prefer conversation'to sequential ex-
change of messages.
2) Duologue and Teleconferencing: Although there has not
ß been, .thus. far, very much use of networks for one-on-one
interaction between users, it seems .likely that some kind of
corn. purer-augmented two-person telephone communication
will one day be one of the main modes of networking. In
order to displace the conventional telephone, "teleduologue"
will probably have to offer speech, writing, drawing, typing,
and possibly some approx/mation to television, all integrated
into a synergic pattern with several kinds of comlter support
and fadlltation. The two communicators (and their supporting
programs) will then be able to control displ,s in certain
areas of each other's display screens and processes in certain
sectors of each other's computers. Throughoi:t a duologue,
each communicator will be advised by his own programs and
will use information from his own data bases and other sources
accessible to him. The effect-will be to provide each com-
municator with a wide choice of media for each component of
his communication and with a very fast and competent sup-
porting staff.
A teleconference [14]-[17] is an organized interaction,
through a communication system or network, of geographically
separated members of a group. The term "teleconference" has
been used recently mainly to refer to interactions organized
or presided over by or with the aid of programmed computers.
In some teleconfercnces, the'members of the group participate
concurrently; in others, each member logs in when it is con-
venient for him to do so, reviews what has happened in h/s
absence, makes his contribution, and logs out, perhaps to re-
turn later in the day or later in the week.
During the last five years, a cousiderable amount of experi-
ence has 'been gained with computer-facil/tated teleconfer-
encing, but it /s evidently a complex .and subtle art, and
teleconference programs still have a long way to go before
teleconfere'nces approach the naturalhess of face-to-face inter-
.action. On the other hand, we note the inefficiency of travel-
ing to meetings and the inefficiency of letting one participant
take up the time of n - '1 participants when only m < n - 1
are interested in. what he is saying. A/teleconferencing is per-
fected (especially nbnconcurrent teleconferencing), it will be-
come an extremely important technique.
C. Neopaperworir
"Office automation," "computer-based officework," "the
high-technology office," and a few other such phrases refer to
the aggregation and integration of several applications'of com-
puters and networks in office work. ("Automation" is in-
tended in its weak sense, which includes computer "'aiding"
and "'semi-automation.") Office automation includes every-
PROCEEDINGS OF THE IEEE, VOL. 66, NO. 11, NOVEMBER
thing presently called "word processing" (dictation, docun
preparation, etc.) plus computer-based filing (informa
storage and retrieval), communication (electronic mail, t
tronic message services, duologue, teleconferencing),
modeling (simulation), and it 'connects with electronic fu
transfer, management information systems, and parts-if
all-of computerized commerce.
Office automation is expected to make heavy use of:
Works, both local and geographically distributed. Much of;
work is organized in an approximately hierarchical manr
with component desk functions such as transcription, editi
f'ging, retrieval, scheduling, and telephone answering at l
echelons and corporate' or divisional functions such as planni:
marketing, operations, and public relations at high echelo
LoW-echelon functions typically are carried out locally, wit
a single office or suite of offices, and, when low-echelon fur
tions are supported by minicomputers or microcompute
local networks will be required in their integration.
geographically distributed organizations, of course, geograp!
cally distributed networks will be required as higher level fur
tions are integrated.
]} Telework: Networks will make it possible for people ß
do informational work effectively at locations remote fro
their managers, their co-workers, the people who report
them, and, indeed, even from customers and clients with who:
they must interact. Such telework will require facilities fc
duologue, teleconferencing, and all the other aspects of offic
automation-but little beyond what automation of a nondi:
persed office will require.
Telework will offer the possibility of saving the hours an½
the energy spent in commuting. It may burden some familie:
with more togetherness than was contracted for through the
marriage vows ("for richer or poorer, but not for lunch"). BuT
its srongest impact on individual lives will surely be felt by
persons immobilized by prolonged illness, physical handicap.
or children. For many of them, networks will open many
doors-including the door to gainful employment.
2J Augmentation of the Intellect: The at-a-distance aspect
of computer-based work that is emphasized by the term "tcle-
work" will be overshadowed, in the opinion of many, by what
Engelbart [18]-[2I] has called "augmentation of the in-
tellect." Computers will help people do informational work
faster and better by providing fast and accurate tools to
supplement such slow and fallible human functions as looking
up words in dictionaries, copying references for citation,
stepping through checklists, and searching for matching pat-
terns. Augmentation is needed at levels that range from
A) helping poor typists who cannot spell to put out neat and
accurate reports, to Z) improving the content and style of top-
level policy statements. Expectations differ concerning the
prospect for significant early contributions from artificial in-
telligcnce, but it is clear that relatively unsophisticated aug-
mentation systems can make major contributions. Consider
the help provided by descriptor-based and citation-index-based
information retrieval systems to a person looking for refer-
ences pertinent to a particular fact or concept. Or consider
the impact that would be made, on writing such as this, by a
text editor that automatically displayed the Flesch Count [22]
of every paragraph it helped compose.
3J Tastr Management and Coordination: In addition tohclp-
ing the /ndividual worker, computers will facilitate teamwork.
Each office task will have its planned course of' actions, in-
------------------------------<page break>-----------------------------
LICKLIDER AND VEZZA: APPLICATIONS OF INFORMATION NETWORKS
1333
volring particular workers at particular projected times. A
computer-based task management system will monitor the
task as it moves along the course, checking the actions as they
" are taken, arranng that planned coordinations and approvals
are obtained, and revising the plan (or calling for human help)
when the schedule slips. In the early days of office automa-
tion, the task management process will be mainly a matter of
maintaining orderly work queues for tl!.e office workers and
displaying for them at each toomen{ 1) what needs to be done
i:. and 2) the information needed in doing it. What will need to
:;!. be done will usually be, of coute, to solve a problem or to
.}' make a decision-nlost of the preliminary work will have been
ß :: performed automatically by computers. With the passage of
' time, as people come to understand the problem-solving and
"- decision-mnklng processes and the supporting. information in
programmers' terms, computers will chip away at the problem-
solving and decksion-making substance of office work, but we
expect the now-rising wave of office automation to succeed or
fail on the measure of its help to human workers and to
hu
D. Management Applications ' ,z
Office automation will have its impact upon managennt,
of course, as well as upon the office workers. Managetment
deals almost exclusively with information. (Money is essentially
information, of course.) The comptrolier's department was
com early. Electronic funds transfer will be a major
application of special-purpose, limited-purpose, or general-
purpose networks. On-line financial services may burgeon. In-
ventors,, ordering, production, pricing, and planning will all be
interrelated with the aid of networks and computer modeling.
1} Management Information Systems: The widespread feel-
ing ß of disappointment in the management information systems
(MI ]'s) [23]-[26] of the 1960's and early 1970's had, we bern
lieve, a simple basis: the activities that generated the informa~
tion required to support management decision making had not
been brought on line to computers, and, therefore, the re-
quired information was not available to the management in-
formation systems. To some extent, information important
to the manager ks so global in scope that capturing it all on
line ks still not possible. (It was not worthwhile to keypunch all
the basic operating data just to feed them into the manage-
ment information system, for only a small fraction of the
totality would ever be used. It was impossible to anticipate
just what subsets or aggregations of the basic operating data
would be required.) As soon as all the informational activities
involved in operating an organization are on line, however, the
basks for an effective management information system will
exist A few organizations are already approaching that state,
but most are just entering-or just beginning to contemplate-
office automation.
Local and geographically distributed networks will make it
possible, at a cost, for top management to access all the facts
and figures involved in the minute-to-minute operations of a
business. Top management should resist the temptation to
convert that possibility into actuality. The principle that
looks best at present ks to let the data of a corporation reside
where the managers most conversant with them reside ("keep
the data near the truth points") and to have conversant man-
agers "sign off" on the release of data upwards in the corpo-
rate tree. Certain data should be abstracted and moved upward
according to preset schedules; other data may be queried from
above-but queried through an authenticating release procesa.
Of course, the release process may in some instances be medi-
ated by programs operating on behalf of the human conversant
manager rather than by the human conversant manager himself
or herself.
The foregoing ctkscussion pertains, indeed, to most of the
data management func, tions in office automation. In dis-
tributed organizations, data will be distributed, and one of the
main uses of networks will be to move data from points of
residence to points of use.
2} Modeling and Simulation: Computer-based modeling and
simulation are applicable to esentially all problem solving and
decision making. At present, however, modeling and simula-
tion are computer applications much more than they are net-
work applications-and they are far from ubiquitous even as
-computer applications.
The trouble at present is that most kinds of modeling and
simulation are much more difficult, expensive, and time con-
suming than intuitive judgment and are cost-effective only
under special conditions that can justify and pay for facilities
and expertise. But those are prime conditions for resource
sharing and, hence, networking. Whereas very large organiza-
tions will be able to a/ford their own concentrations of facili-
ties and expertise, small organizations will not. A management
grows tighter and more sophisticated, therefore, there may
come to be a place for management consultation and service
lb-ms that specialize in modeling and simulation and offer very
large or special facilities-and deliver their products through
networks. Perhaps a glimpse of such a future has been given
by the large array-processing computer, Illiac IV [27], [28],
which has been used through the ARPANET in modeling the
world climate and the space shuttle. Similarly, MIND [29], a
system accessible through a value added packet network, is
being used to design communication networks.
E. Commerce
Shifting our attention from activities within an organization,
such as a business firm, to interactions among organizations,
we can see another kind of application for networks.
2} Electronic Markets: Networks will serve as marketplaces,
providing meeting grounds for buyers and sellers. At first,
networks will dksplace telephone and mail, which now serve
the marketplace function for most businesses. Later, networks
will begin to displace stock exchanges and commodity markets.
Ordinary office automation and funds transfer facilities will
adequately support negotiations and transactions when the
"commodities" bought and sold are purely informational or
sufficiently specifiable by words and figures. Wide-band facili-
ties for examining products at the time of purchase ("squeezing
the grapefruit") may extend the scope of the electronic market-
place to commodities that must be selected or approved in-
dividually by prospective purchasers. We can expect networks
to go beyond the role of the mere place or medium for trans-
actions and, with the aid of sophisticated programs, actively
to "make a market" in the sense that certain stock brokers
make markets in certain stocks.
2} Computerized Commerce: Computerized commerce [30]
is based: on the idea of electronic markets. It goes beyond pro-
riding a marketplace and making a market-and back into the
primary motivation of the business firm-by using computers
to develop and carry out the strategies and tactics of buying
and selling. The concept of computerized commerce is appli-
------------------------------<page break>-----------------------------
0ourse, be available to supplement the computer knowledge
but considerations of cost and availability will almost
favor the computer. Difficult problems of legal re-
and liability may have to be solved: advice from a
base may be similar to advice from a book, but a
knowledge-based program that controls the administration of
an anesthetic would appear to introduce a new factor.
Possibly even more far reaching in its implications than
access to medical knowledge bases by physicshe/s access to
medical knowledge bases by ,laymen. Knowledge bases for
hymen would have to be quite different in content and pack-
aging from knowledge bases for phys/cians, and ideMIy the
two applications would complement each other. The layman-
oriented application might deal maRfly with the complaints
not ordinarily taken to a doctor or with the dec/sion process
that determines whether or not to seek a physician's help. In
either case, ff the knowledge-base program had access to the
individual's medical record, and ff it could make simple
observations such as temperature and pulse rate through the net-
work, it could go rather far beyond the Emits of the conven-
tional book of medicine for the layman. Society should ex-
,i
amine such incumions by the computer into medicine r even
paramedicinc very carefully before making up its c11ective
mind about them. They obviously mix benefits with dangers.
Unfortunately, they tend to he approached with prejudice.
full U . . ß '.',":',.:4,
works. For reasons we do not Y nderstand-sme:.).
response to a changing situation is the essence of com4''f"
and control-the World-Wide Military Command and' Control
System (WWMCCS)' is actually not very interactive, andJit!'
computers, which use the GECOS operating system [36]' de;
signed for batch processing, are not/nterconnected. by an. dec-.
tronic network. But surely vtrMCCS wil//n due course be, up-
graded. Autodin II is under development and wil/supplement
or replace Autodin I [37], the Department of Defense's present.
store-and-forward digital telecommunications network, with a
modern packet-switching network based on modified and
secured ARPANET technology. Networking is be/rig pursu&t
actively, also, in the intelligence community. One of the .most
significant possib/I/ties for the military that ls opened ui by
advances in /nformation technology /s the achievement of. a
much tighter coupling between intelligence and command and
control. One can envision a reduction in the time required for
the distribution of intelligence information from days or hours
to minutes or seconds. Such an advance would, of come, put
pressure on intelligence gathering and processing to operate on
faster time scales.
2) Military Logistics: There is less progress, but also less
pressure, in the logistics area, where more than 20 large batch'
inventory systems can be counted, diverse in respect of hard-
ware, programming language, and datg management systerd.
Over the coming years, however, even the logistic, situation
will probably be brought under control and onto a network.
The overall objective/s to make the enrire operation of a
tary effort responsive to coherent hierarchical command in the
light of valid and current intelligence-with security
enemy actions and countermeasures.
3) The Network of the National Crime Informetlon Center'
{NCIC): The NCIC is operated by the FBI and connects Vith
state and local police units in most of the states, The NCIC
contains, among other things, data on stolen cars and stolen
1/cense plates and the police histories of convicted criminals.
The case of the NCIC network is an interesting study because,
in it, the informational needs of the police and the information-
providing capabilities of computers and telecommunications
mn head-on into Congressional concern for the right of in-
formational privacy. When a police officer stops a speedflag
car and approaches it to make an arrest, he would like to know
something about the car and driver. Is the car stolen? Does
the owner have a h/story of res/sting arrest? Forewarned is
forearmed. About two years ago, however, an innocent man
was killed by an arresting officer forearmed by forewarning
with incorrect information. In the most recent chapter, the
Senate Committee on Government Operations caused to be
rejected the NCIC's request for perm/ss/on to acquire a message-
switched network to speed up commurdcation with state and
local police.
4) Social Security: In the U,S., the Social Security Adminis-
tration (SSA) distributes more than $100 billion a year to
more than 20 million people and interacts with millions of
clients each year through about 1300 offices manned full-time
and 3000 manned part-time. The set of computer processible
data bases that support SSA operations contains more than a
trillion characters, and it is estimated that in those operations
.-e..:7: ¾e*_r several trillions of characters flow from' one location
t,: :v. ork.'..:. .:....u  tenth of a trillion by a network and process-
i:,:::. :vste:;..:.'.':--.3 ''e ' ::.3  DARS system," and the ret mainly
G. Government Applications
Actual and potential government applications of networks
include military command and control, communications,
logistics, acquisition and interpretation of intelligence data,
dissemination of intelligence, law enforcement, delivery of
government services such as Social Security benefits to citizens,
.End converting the paperwork of the bureaucracy into bits.
Paperwork in the government/s rather like paperwork in the
private sector, but earned a step or two further into detail.
The other government applications, on the other hand, seem
- rather special. Military command and control, communica-
tions, intelligence, and to a cons/derable extent logistics sys-
tems must be able to operate fast, move fast or hide, and
function in the presence of physical (as well as other) counter-
measures. Law enforcement information systems are in some
ways like highly ampiflied credit reference systems: derrogatory
information seems especially crucial, for to be forewarned is to
be forearmed, and action must often be taken on the basis of
whatever data can be assembled in a few seconds, Serving
the citizens and collectg taxes from most of them requires
that certain personal data be held about almost everyone-
enough in sum to make a several-trillion-character data base
that at least conceivably could be subverted to political or
economic exploitation. There are strong lessons about govern-
ment applications of networks in the recent rejection by the
Office of Management and the Budget (OMB), at the well-
timed suggestion of several members of Congress, of the pro-
posed new Tax Adm/nistration System of the Internal Revenue
Service.
]) Military Command and Control and Military Communica-
tions: Military command and control and military commun/-
eot!?.-...: ".7'. :z.:'.'..; .:-'.';'c.'!: '-,,'eh'eat;ons. Both interactive
5e_t½ ,:,'_';?j-A.Z.'C:'r!.,r!c '3:':2..2.; -,;i'."g"',;',zZt :"?'.' :"r
all(l 2;a.;sv u'f :':,_6 :112!'.Z:'x. i.;':i½rr, s' U&3/ [o '2:'"2.'_1 .""- 't:i'e;.œ
------------------------------<page break>-----------------------------
1336
PROCEEDINGS OF THE IEEE, VOL. 66, ,NO. 11, NOVEMBER 1978
In 1976, the SSA began planning the modernization of the
process through which it discharges its massive responsibilities
[38]. The new process will make even heavier use of com-
pgters than does the present one and there will be much con-
sultation and updating of central or regional data bases from
local offices. (The present process requires several computer
areas, each with multiple mainframes, more than a hundred
disk drives, and more than a hundred tape drives-and, in all,
approximately 400 000 magnetic tapes.) The new communica-
'tion subsystem will, therefore, be a network of very major
propbrtions, probably a dedicated SSA network operated by
the General Services Administration or (improbably in the near
term) a part of an even larger and more comprehensive net-
work. Most technologically developed countries will sooner
Or later have social security networks.
, ff. Protection
If military, intelligence, and police networks are reckoned
as networks for protection, then protection is a very large
category of network applications. There is another member
of the class that deserves mention.
lJ Home and Neighborhood Security: Several of hc pro-
jected applications of computers in the home relate to,security:
sentry against intrusion, fire, and gas and water leakage, mord-
toring the well-being of the elderly and infirm, and "electronic
babysitting." In most of these applications, computers will
be better at detecting trouble than in correcting it, and there
will be a strong requirement for communication with remote
persons or agencies. At present, some burglar alarms are con-
ncctcd by dedicated lines to central security offices or police
stations and some "dial up"in the event of trouble. If a packet
ß network were available, it would probably be less expensive
" and it would provide a wider range of options, including absent
ß members of the family, friends, and neighbors as well as
curity cornpan/cs and public agencies.
It sccrns possible that a neighborhood communication
medium (with a broader fan-ot or faster sequencing of calls
than the telephone) might bc just what s required for the
elderly to help one another achieve a higher level of security
and peace of mind. CB radio or house-to-house (or apartment-
to-apartment) wiring or a multipurpose packet network could
provide the medium.
Probably just conversation of the kind that prevails on CB
radio interconnections would go a long way, but it could be
reinforced by slightly higher technology. Home computers
could be programmed to interpret a variety of indicators of
trouble-the sound of a fall, too long a flow of water, the re-
frigerator door open, prolonged quiescence-and to ask for an
"all's well" report whenever there was cause for concern. Fail-
ing to be satisfied that all was indeed well, the computer could
cali for help. It would have a list of participating neighbors
and a schedule of probable availability for each, ß and-by com-
municating with their home computers-it could quickly find
someone to look in and check, or provide assistance. It has
been suggested that a neighborhood net could monitor its
clients while they were walk/ng on the sidewalks as well as
while they were at home, and a small device has been demon-
strated that sends out a radio signal when its wearer falls
down-or for some other reason becomes horizontal [31 ]. If
neighborho6d networks existed, there would probably be no
end of inventions to exploit them in the interest of security:
heartbeat monitors, breathing monitors, footstep sensors, and
so on. And if the present trend of population statistics con-
finues, security applications might constitute a significant
sector of the network application pie.
I. Education and Awareness
Beginning with the last section and continuing now into this
one, the focus of interest has moved from the organization-or
the indiv/dual as a member of an organization-to the in-
dividual as an individual in the primary family group in the
home. Probably the most important network prospects for
the individual in tkis century lie in education and training.
1} Computer-Based E'ducat'on and Training: We assume
that advances in computer representation of knowledge anin
computer mediation of interactions between people and
knowledge bases will advance computer-based education and
training far beyond the "expensive page turners" and drill and
practice routines that are associated in many mind with the
term "computer assisted instruction." We assume that knowl-
edge bases accessible through networks will eventually accumu-
late more knowledge, in each of many fields of learrdng, than
typical teachers are able to master and retain, and that the
knowledge in the knowledge bases will be well organized (by
experts in each field) and effectively accessible to students at
all levels of mastery and aptitude. However, computer-based
techniques for the representation, organization, and exploration
of knowledge are at present still topics of research-and even if
they were fully developed today, it would still take a decade or
two to translate the content of the many fields of learnhug into
computer-pmcessible knowledge bases. During the coming
years, therefore, application of networks in the area of
computer-based education and training will be preliminary and
propaedeutic. Perhaps toward the end of the century it will
approach its ultimate volume and significance-and be among
the top three or four uses of networks.
2} News: At present, most people gain their awareness of
what is going on in the world mainly through mass media that
report on events rather than processes, that select a few news
items instead of covering the news, and that give everyone,
regardless of his or her interest pattern, the same few selec-
lions. Networkhag has the potential of changing the news into
a multidimensional dynamic model of the world that each in-
dividual can explore in his own way, selecting for himself the
topics, the time s'cales, the levels of depth and detail, and the
modes of interrogation and presentation. Interest profes and
other techniques of selective dissemination may play im-
portant roles, but networking in principle removes the neces-
sity of dissen/mating (with its implication that the initiative
I/es mainly with the transmitter) and opens the door to self-
directed exploration and investigation by the receiver of the
news. To provide the multidimensional.dynamic model for
exploration and investigation would, of course, be a demand-
ing responsib/lity for the gatherers and organizers of news, but
they gather and organize much more even now than they print
or (especially) broadcast. There will probably be a long slow
evolution from the newspaper/newsmagazine format and the
nightly news format through increasing levels of user initiative
toward truly user-dominated interaction with a whole-world
knowledge base.
III. RV. QmRV. MV. NTS IMPOSEr) UPON NETWORKS
BY APPLICATIONS
Now that we have sketched out several applications, we
should examine briefly the network characteristics they re-
quire. The applications do not all require the same network
characteristics, of course. One. application may require one
------------------------------<page break>-----------------------------
AND VEZZA: APPLICATIONS OF INFORM.TION NETWORKS
133'7
of characteti, wh/lc s. aothcr application may re-
another pattern.. Some of the frcqucnfiy requ/red char-
ristics are the following.
Bidirec'orl Transmission: Most applications requ/re
communication-if not the capability of sending and
simultaneously (fuLl duplex), then at least the cape-
ß b/1/.ty of alternating between sending and receiving (half duplex).
2) Freedom from Error: One wrong bit may completely
'ehange the meaning, especially if numerical data are repre-
sented nonxedundantly. In such cases, even though the basic
channel itself is rot error free, the end-to-end
must be made error free through the use of
':'adequate error handling mechanisms.
ß ? 5) Efficiency Despite Burstiness: A source that transmits
?' short bursts of information and is quiescent between bursts
typically does not wish to pay for channel time wh/le it is
quiescent. Both human behags and computers are bursty
soutees.
4} Low Cost per Bit: Te cost of network service depends,
of course, upon many nontechnical factors as well a upon the
tec]xical efficiency of the network in converting its resources
into services. But technical efficiency is a very strong and
basic factor. This characteristic refers to the cost of tns-
mirting one bit from source to destination. The telarisen of
cost to distance is considered separately.
5} High Connectivity: A source may need to transmit. to any
one or more of many destinations. A destination (i.e., user)
may need to examine many sources.
6} High rnformation Rate: Wide-band charmeis are capable
of transmitting many bits per second. The criteria for
"wide," and "many" vary widely with type of signal and level
of expectation. The 50 000-bits/s irfformation rate of most of
the ARPANET channels seems like a hJg information rate for
odinary interactive computing, but it is too low for con-
venient transmission of large files (e.g., ffig.h-resolution photo-
graphs) and far too low for m oving pictures, even low-resolution'
television pictures.
7) Security: Security is a comItex of characteristics, some
of which provide the techrdcal basis for the protection of
privacy. Others have to do with preventing disruption of
senrice and protecting against fraud and theft.
8} t%'vacy: In the U.S., is complex of informational
rights, including but by no means ]j_rrfited to protection against
eavesdropping, has been formuiated by the President's Com-
rrdssion on Privacy and, to a considerable extent, expressed in
legislation in the Privacy Act of 1977. Other nations also have
privacy laws, of course, some of them in some respects more
stringent Man ours.
9} Authentication.' A good authentication scheme provides
the electronic equivalent of a signature. Ideally, authentication
identifies the author of a document and'makes it impossible
for m to escape responsibility for the authorship. Ideally,
also, authentication makes it impossible for anyone to change
even one character or bit of the document without destroying
the "signature."
J O} High Reliability: Low probability that network service,
as seen by the application, will be impaired by macroscopic
malfunctions. For present purposes, we distinguish between
macroscopic malfunctions and microscopic errors in bit
transmission.
J J} Full-Duplex Transmission.' Some applications require,
and most axe favored by, the capability of sending to another
station and receiving from it at the sarae time.
12] Priority Service: Guaranteed or preferential service,
pccially when the network is congested, is widely regarded as
essential for certain very important functions or for certain
very important persons.
13) Speech Capability: Present speech circuits transmit
alphanumeric information inefficiently, and most present data
networks were not designed to transmit speech. It will be ad-
vantageous, however, to integrate speech with data.
14} Pictures: It will be advantageous to integrate pictures,
also, into the repertoire. Graphs, charts, diagrams, and simple
sketches fit readily into the pattern of data transmission, but
high-resolution pictures and, especially, moving pictures re-
quire high information transmission rates. fis characteristic
is essentially a second "information rate"-but scaled in such a
way as to be more demanding of very-wide-band capabil/ties.
15} Insensitivity to Distance: Synchronous satellites and
packet switching both tend to make the difficulty and cost of
transmission less dependent on distance than they are in tradi-
tional communication systems. Rarely is it an absolute re-
quirement that difficulty and cost be independent of distance,
but often it is desirable.
J6} Short Transit Time Delay: If the sum of the signal-
transit time and the signal-waiting-in-buffer time is too great,
an application may be slowed down too much or disrupted.
The 0.2-s delay introduced by transmission via a synchronous
satellite somewhat disturbs two-way speec h communication.
The delay introduced by transmission from one processor to
another may slow down the operation of a multiprocessor that
is a network of minicomputers or microcomputers.
1 7} Uniform Time Delay: In some applications, successive
segments of the signal must reach the destination in sequence
(or be put back into sequence if they arrive in scrambled
order). Note that reordering may cause all the segments to be
delayed as much as the most-delayed segment.
18J Broadcast Capability: Some applications require, and
some are favored by, the capability of transmitting to many or
all destinations concurrently.
J9) Mobility: Some or all of the stations may need to move
from place to place and may need to communicate in transit.
To obtain rough measures of the requirements imposed upon
networks by the several applications, we feed in the body of
(an early version of) Table I. i Into each cell we entered a
number to indicate our intuitive rating of the importance qf
the characteristic for the application. The rating scale we used
runs from 0 (lowest) to 5 (highest). For example, we con-
sidered connectivity to rate at 4 in importance for mail and
message systems because mail and messages typically fan out
widely from senders to receivers and fan in to receivers from a
wide distribution of senders. We did not assign a 5 because
mail and message systems would still be valuable (cf., the plans
of Satellite 'Business Systems) if connectivity were limited to
within organizations. In the case of the column 6, information
rate, we used a somewhat special scheme. The numbers from
1 through 5 encode five class intervals of information rate in
 In order to obtain a broader basis on which to think about, and
possibly model, the relations between networks and applications, we
suggest that you (the reader) photocopy Table I and f'fll in some rows
with your estimates of the importance of the characteristics to the ap-
plications. We have also provided room for you to define additional
applications or characteristics. If you are willing to share your estimates
with us, despite the fact we are not bold enough to share our raw-data
estimates with you, please post them to J. C. R. Licklider and A. Vezza,
MIT-LCS Rm. 219,545 Tech Sq., Cambridge, MA 02139.
------------------------------<page break>-----------------------------
1338 PROCEEDI2GS OF THE IEEE VOL. 66, NO. 11, NOVEMBER
TABLE I
IMPORTANCE COEFFICIENTS OF NETWORK CHARACTERISTICS FOR VARIOUS CLASSE OF NE'TWORI APPLICATIONS (EACH CELL ENTRY SHOULD BE
SUBCTIVœ RATING OF IMPORTANCE ON A SCAI- IOM 0 (Low) TO 5 (HiGh))
NETWORK
CHARACTERISTICS
APPLICATIONS
COMMUNICATION
NEOPAPERWORK
MANAGEMENT
COMMERCE
PROFESSIONAL
GOVERNMENT
PROTECTION
EDUCATION AN AWARENESS
SPACE FOR ADDITIONAL
APPLICATIONS
------------------------------<page break>-----------------------------
 ER AND %rEZZA: APPLICATIONS OF INFOIVIATION NETWORKS
," RANK IN IMPORTANCE
AMONG 19 CHARACTERISTICS
t IMPORTANCE COEFFICIENTS OF 19 NETWORK
100 CHARACTERISTICS'
o 80
 70
60 .
40
30
- 20
lO
0
o ,,, >. 
1339
Z
5O
45
40
35
30
DEMANDINGNESS COEFFICIENTS OF 30 NETVVORK
APPLICATIONS '
10
5
o
RANK IN DEMANDINGNESS Z
AMONG 30 APPLICATIONS
Fig. I. Importance Coefficients
24261916 216 6 8 52116 91230 1212 2 326 19289 9 I 18324 28 1521
of 19 network characteristics and demandinghess coefficients of 30 network applications. (*Divide each num-
ber by 1000 to scale the sum to unity.)
;.bits/s: 1) 75-300, 2) 300-1000, 3) 1000-10 000, 4) 10 000-
1 000 000, and 5) above 1 000 000. For the rough purposes
of our analysis, nevertheless, we shall interpret the entries in
.. column 6, as all the other entities, to be estimates of the im-
Portance of _the (columnar) characteristic for the (row)
application.
Fig. 1 shows the relative importance of the 19 network char-
acteristics. As one examines the average ratings of the charac-
teristics, it comes as no surprise that bidirectionality is very
important. It is the "co" in "communication."
It may be slightly surprising, however, that freedom from
error is so important. It is freedom from error as seen by the
------------------------------<page break>-----------------------------
1.340 '., :.
Of 0'. e  bound tO b= eo  the aw
'½network' channels, but they may be detected and eliminated by
;'-',-; error-correcting circuits or by retransmission. "Error-free"
" .may in practice mean one bit error in 1012 or 1014 bits, on
the average. The importance of achieving an extremely low
'. ', error rate stems in part from the fact that many of the applica-
!"- ' tions involve information, such as financial data,,in which
changing a single character could make a great difference.
"Freedom from error is 'required, also, by most cryptographic
'?" schemes Where freedom from !trot is not required by an
application, one can usually fiffd error detecting and correcting
mechanisms within. the application itself. Such mechanisms
'. are quite evident, for example, in human conversation. But it
:::(. greafiy: simplifies most network applications if the network
can be counted on to do the error handling.
Ability to handle bursty transmissions efficienfiy ranks third.
The advantage provided by-this characteristic translates di-
rectly into a cost advantage.
c."' Low cost ranks fourth in importance. It did not rank higher
; because we recognized that certain of the applications, such as
military command, control, and communication, are relativ .el.y
:." insensitive to cost. Also, other network applications suclk/a
mail and messages are already quite cost competitive with th%ir
conventional counterparts and do not demand very-low-cost
facilities.
Connectivity ranks fifth in importance. The reason con-
nectivity does not rank higher is that we assigned only a
medium score for connectivity to applications that required
only connectivity within an organization or within a region,
and many applications could function-though perhaps at
: some disadvantage-with such limited connectivity.
Information rate ranks sixth. We interpret that to mean that
very .'ide-band transmission/s not vital to most of the applica-
tions and that most of them could be satisfied with an in-
formation rate in the range 1000-10 000 bits/s. However, that
is the information rate seen by the application. To handle
heavy traffic, and to handle a few cf. the applications, a net-
work should have channels of considerably greater bandwidth
than that.
Security, the complex that includes assurance of service
when required and protection against fraud and theft, ranks
seventh.
Privacy, the complex that includes protection against dis-
closure of personal information and unauthorized use of it,
ranks eighth.
Authentication, a characteristic closely associated with se-
curity, ranks ninth.
, At the other end of the ranking, mobility (19th) is not re-
' quired by most of the claases of applications we considered-
but, of course, is essehtial for some applications..
Broadcast capability (18th) was scored low. because it is not
needed at all in many applications and is needed. only
caslonally in others such as mail and message systems-and,
when needed, usually can be simulated adequately by repeated
point-to-point transmissions.
Uniformity of time delay (l?th) is important mainly for
'speech transmission. If speech had been given a we/ghting pro-
.. porttonal to its probable eventual importance in networking,
uniformity of t/me delay would have ranked higher.
The capab/i/ty'of giving preferential treatment to kigh
priority traffic (12th) ranks as high as it does because we
..:'._viewed priority in. the context of present-day systems that
ay introduce .considerable delays into the delivery of some
of their messages. In the context of future systems in
PROCEEDINGS OF THE rEEE, VOL. 66, NO. 11, NOVEMBER 1978
which a whole transmission will take less than a second,
priority may be much less important. However, the need for'
priority is ullkely to-wanish. Priority classes are useful in
queuing messages for processing by people and in indicating
the ptioritizer's sense of urgency or importance to the re-
cipient. Moreover, even very wide-band systems tend to be
designed just barely to handle expected peak loads, and even'
such systems can be overloaded-in which case, ptiotitization
might be helpful. On the other hand, it is conceivable that the
processing of ptiotities might slow a system down more than
eliminating low priority traffic could speed it up.
We do not want to attribute too much value to our no-
doubt-ideosyncratic subjective estimates of the importance of
network characteristics to applications, but we would Eke to
carry the analysis another step to illustrate what we think
might be a valuable method. It attempts to deal with the tela-
live merits of various networks or network architectures.
The method begins with a table of applications versus char-
acterist:ics similar to Table I except that each application has
an importance weight and all its cell values are multiplied by
that weight before the colunms are totaled. The method as-
sumes, also, a table of networks or network architectures versus
network characteristics such as Table II. The entries in Table II
represent our very subjective impressions of the degrees to
which the characteristics at the top characterize the networks
at the left-hand side. The values are certainly not definitive. In
the case of the hypothetical augmentation of the ARPANET,
they assume major increases in number of subscribers and in
information rate, and they assume that advanced provisions are
made for scurity, privacy, authentication, and ptiotity service.
They assume, also, that satellite relays are incorporated into the
network along with wide-band surface channels and that there
is a packet-radio subsystem to serve mobile applications.
In the case of the projected SBS service, indeed, they are
based only on the most informal information, and they make
rather optimistic estimates about the characteristics of the
hypothetical networks that would be developed on the basis
of the SBS facilities. The reader is invited to substitute his or
her own estimates.
To determine the suitability of a network or network archi-
tecture to a set of applications, one simply multiplies each ceil
value in its row in (the table like) Table II by the importance
of the corresponding characteristic at the bottom of (the table
like) Table I-and then finds the sum (across the row) of the
products.
To illustrate the use of the method, we worked with the four
networks of Table II and with four sets of applications. Appli-
cation set 1 was the set shown in Table I. Sets 2, 3, and 4 were
subsets consisting of-set 2: speech and encrypted speech,
set 3: still and moving pictures, and set 4:. duologue and aug-
mentation. (We gave equal or uniform weighting to the appli-
cations in each set-to all 30 in the first set and to both of the
twoin each of sets 2, 3, and 4.)
We obtained the 16 appropriateness indexes shown in Table
III. The dial telephone network performs best in the speech
apphcations, of course, but it does not appear to do badly in
the others. (Giving more weight to cost tends to reduce its
scores.) The experimental ARPANET appears'to perform well
on speech, which surprised us despite the fact that experiments
on the transmission of compressed speech over the ARPANET
have been very successful, but best on duologue and augmenta-
tion, which we expected. In the scoring, the ARPANET suffers
Because, being experimental, it was not developed in respect to
some of the important network characteristics. Assuming such
------------------------------<page break>-----------------------------
AND VEZZA: APPLICATIONS oF INFORMATION NETWORKS 1341
'-,- TABLE I1
Es'rIM^TED DGREES TO WHICH FOUR SELœCTED NETWORKS POSSESS TIlE 19 NœTWORK CHARACTERISTICS
(THE RATtNGS ARœ TIlœ AUTHOœS' INTUtTtVœ ESTtMATES ON A SCALE FROM 0 (Low) TO $ (HtoH)
 NETWORK I 2 3 4 :5 !6 '7 8 9 10 11 12113 14 15 16 117 18 19
Dial Telephone 5 2 0 2 5 3 2 2 1 3 2 1 4 2 0 4 4 2 2
ARPANET 5 4 4 3 3 3 1 2 1 4 4 0 21 21 4 3 3 '1 1
Hypothetical Augmented
ARPANET' 5 4 4 4 5  5 4 4 4 4 4 3 4 4 4 4 4 3' 4
Hypothetical Corporate
Network UsingSBS" 5 4 4 4 2 5 4 4 4 4 4 3 4 5 4 2 4 3 0
*A hypothetical network based on ARPANET technology but with very wide-band ground and satellite chan-
I_ nets very many subscribers, advanced provisions for securdy. an authentication scheme. arrange-
ments for priority, and a mobileJportable radio adjunct 'based on the ARPA Radio Net,
ß ''A hypothetical network of the kind Ihat might be based on the protected facilities of the Satellite Business
Syslems Corporalion and used by a large corporalion with geographically distobuted branches. It rs assumed
i' that this network is used only within the corporation and therefore has restricted connectivity,
/j.. TABLE III
APPROPRIATENESS SCORES FOR FOUR SELECTED NETWORKS, EACH RATED
,'': o Fou DtRENT SETS OS AVV.IC^ONs (Ts AVROVm^TENESS
,_ ;' S, CORES ARE BASED ON A SCALE FROM 0 (Low) TO 5 (HIGH).
B;- THeE WAY THEY WERE DETERMINED IS DESCRIBED 1N THE TEXT.)
1 2 3 4
APPLICATION ½
'- 
o "' ' :
Dal Telephone 2.5 2.8 2.4 2.6
Expedmentat ARPANET 2.8 2.8 2.5 3.0
Hypothetical Augmented
ARPANET , 4.1 3.9 4,0 3.8
Hypothetical Corporate
Net Based on SBS Seice 3,8 3.5 3.9 3.8
development (Hypothetical Augmented ARPANET) yields the
appropriateness scores in the third row of Table III, which are
all toward the upper end of the five-point scale. The hypo-
'.thetical corporate' network based on the projected SBS service
to perform almost as well as the hypothetically aug-
:,.mented ARPANET, suffering in the comparison only because
:/we assumed for it limited connectivity (intercorporate com-
"munication only), no surface channels (and therefore always
the 0.2-s satellite delay), and no mobil/t),. Those lacks showed
up only in the average over the 30 applications and in the
Speech applications.
Obviously, the result obtained with the method is no better
than the ratings it processes, and we do not make any claims
for our ratings. We believe, nonetheless, that the scheme puts
into an orderly array some of the basic factors that determine
the relative appropriatenesses of various networks for various
sets of applications, and that it leads the users of the method-
or at least it led us, as we used it-to consider the factors care-
fully and to think about how they act and interact. Upon
examining the interactions, it soon becomes clear that the
linear weighting scheme smooths over many nonlinear logical
interactions, and that a more advanced modeLwould have to
be more like a computer program than three tables and a
pocket calculator. Nevertheless, the fn-st step has to be to
survey the variables that are active in networking, and the
simple scheme provides a start on that.
IV. ISSUES
.4. Brittleness
Brittleness is approximately the inverse of the lauded com-
plex of system attributes: flexibility, robustness, and graceful-
ness of degradation. Brittleness often arises from a quest for
efficiency or economy. If you space the pony express posts
as far apart as a fresh pony can run, then the marl does not get
through when an emergency forces you to use a tired pony. A
socioeconomic unit with a minLmum-capacitance supply sys-
tem avoids the waste inherent in having products stagnate in
the pipeline but crumbles in a siege. Networks will almost
surely be more efficient than the systems they ßsupplant.
Should not that expectation prompt us to ask whether they
will also be more brittle?
From an engineering point of view, the preferred approach
is to avoid brittleness through judicious choices in the archi-
tecture and design of networks. The dynamic routing feature
------------------------------<page break>-----------------------------
1342  ' PROCEEDINGS OF THE IEEE, VOL. 66, NO, 11, NOVEMBEI 1978
of the ARPANET, for example, permits the network to con-
tinue to operate, as far as two functionally connected host
computers are concerned, as lens as there is some path left in-
tact between them. Indeed, just as it is to reliability, re-
dundancy is the main enneerin$ antidote to brittleness in
networks-redundancy of interconnection, redundancy of
power supply, and redundancy of information storage. More-
over, sophisticated uses of redundancy, as in restructurehie
logic and error-detecting codes, provide much greater returns
in ,robustness than do brute force applications that cost the
SKilie.
B. Electronic Imperialism,or Technology Transfer to the World
Aviation may have had more impact in technologically not-
yet-developed but degeloping countries with poor roads and
few rails than in technologically developed countries that al-
ready had working transportation. systems before airmail
routes and airlines came upon the scene. Brailia, for example,
would not have been feasible without the DC-3. Analogously,
networks may have their most dramatic effects where there are
few critical masses of knowledge and few self-reinforcing
centers of intellectual activity. Networks may link the gee-
graphically separated subcritical loci of cognition in the
veloping world with the concentrated supercri.tical nters of
the developed world, bringing the former deeply intce inter-
action patterns of the latter and making it much easier for the
former to grow and advance. If networks do turn .out to have
such an effect, it may represent a new dimension'of imperial-
ism, or it may open up broad new avenues of technology
transfer, or, as seems most likely, it may look one way to some
and the other way to others.
Although technology transfer to the developing world is
often viewed as a matter of delivering journals, reports, and
books and of transmitting data to third-world countries, truly
effective transfer is a transfer not of data or of information
: but of knowledge and it flows through human interact/on.
ß Perhaps .the most effective pattern involves graduate study by
promising young people from developing countries in leading
graduate schools in developed. countries-followed by work
experience in the developed countries and then return to form
loci of advanced technology in the developing countries. But
the difficulties with even that pattern are well known: reluc-
tance So return resulting in the "brain drain," or isolation after
return resulting in unhappiness and ineffectuality. What is
needed is a way to return without breaking the ties of inter-
action with teachers, fellow students, managers, and co-
workers in the centers of technology that, thanks to their
highly developed intellectual and motivational supports, made
possible the transfer of knowledge ha the frst place. Networks
can fill that need. The ARPANET has provided several
instances, albeit just in the U.S., in which students have re-
rosined functionally and motivationally ½ the research groups
in which they worked for their degrees until they could build
up self-sustaining loci of their own ha theix postdoctoral loca-
tions, It seems very likely that the postdoctoral locations
could as well be in foreign countries, even countries with little
technology, if only they had network connections and suffi-
cient funding tc keep local terminals in reliable operation and
to pay for computing time and storage somewhere on ,the net.
Indeed, it seems likely that technology transfer to developing
countries could become real and effective through no more
than informal extension of patterns of interaction that have
become well established in the ARPANET community. But
there is no reason to keep the 'patterns entirely informal.
Formal associations between univers/ties, not-for-profit
organizations, and business Eu-ms in the have-technology and
the have-not-technology countries would surely increase the
productiveness 'of the technology-transfer enterprise [39 ].
We do not want to try to take sides with respect to, and we
cannot hope to offer a solution to the problem suggested by
the juxtapos/tion of "electronic imperialism" and "technology
transfer." We do suggest, however, that the prospect of net-
working between the developed and the developing worlds
deserves very serious study. From the point of view of the
imperialist, it may well be that packet networks will be to the
not-far-distant future what clipper ships (or were they packet
ships?) and clipper aircraft were to the not-far-distant past.
From the point of view of the countries that need and want
technology transfer, packet links to technologically developed
countries may be by far the best way to get it if they can
figure out how to keep the electronic colonialism from coming
with it.
C. Unity, Federan'on, or Fragrnentaffon
If we could look in on the future at, say, the year 2000,
would we see a unity, a federation, or a fragmentation? That
is: would we see a single multipurpose network encompassing
all applications and serving everyone? Or a more or less
coherent system of intercommunicating networks? Or an
incoherent assortment of isolated n6ncommurdcating net-
works, most of them dedicated to single functions or serdr/g
single organizations? The first alternative-the strongly unified
network-seems improbable: of almost zero probability if the
scope is taken as world-wide and still of very low probabili
even if the scope is taken to be the U.S., which seems to
engender pluralistic solutions to most problems. The third
alternative-many separate noninterconnecting networks-is
what would be reached by proceeding with a plan and, there-
fore, may be judged rather probable, but it would be very dis-
appointing to all those who hope that the whole will be much
greater than the sum of the parts, i.e., that many of the pro-
jected applications will facilitate and contribute to one
another to such an extent that the overall value will grow
combinatorially. The middle alternative-the more or less
coherent network of networks-appears to have a fair prob-
ability and also to be desirable, but it brings with it the
problem of how to achieve enough coherence to support fast
and facile intercommunications among the subnetworks when
required, and that may be a difficult problem. Let us consider
first why coherence is desirable and then turn to the difficulty
of achieving it.
"Coherence" characterizes a system in vhich all the parts
articulate well and function in synergy and in which the sub-
systems are compatible and cooperati.v.½.:..In'i an information
network, coherence is desirable p 'artl: for the same reason it
is deskable in a telephone system( tke'v_alue to a typical user
increases as the number of otheradd. esible users increases.
This is true no matter whether tl6 :offsets are people or com-
puters. The importance of coherence. is amplified, however,
works will involve several'"ifUn.c.,ns.:'operating on common
information. Planning a trip.f0r;*xample, will require inter-
action among: YoU r caleiid' ........ the calendar pro-
grams of people you 'reServations programs of
airlines, car rental }, the funds transfer sys-
tems of your b fl..banks, your company's
------------------------------<page break>-----------------------------
AND VEZZA: APPLICATIONS OF INFORMATION NETWORKS
1343
and perhaps data bases pertinent t bti,es to
on the trip. The computers could .not be of
/n the pl,nn if each function or service had its
)arate network or if their networks were physically
but incompatible at various levels of proto-
coherent interconnection of diverse functions will
if networks are to live up to expectations in elco-
message services, computerized commerce, delivery of
serices that involve both federal'and state governments,
military and intelligence applications.
however, is a condition that has to be planned
for. It does not arise in a short time through evolu-
least not through evolution that conforms to the
-naturai-selection model. Will the
that are operating in the present network dtuation
a sufficient degree of coherence for networks to fulf'fil
promise? On the podtire side is the fact that a standard
Ilacket-switching interface protocol X.25 [40] .was formulated
lnd agreed upon in an unusually short time and that the inter-
cbCannnection of such diss/milar networks as Telenet [3] and the
adian data network has already occurred. Also on the
i)ositive side is the possSbL!ityl that one or two commercial
Z'etworks, such.as the one be/rig developed by AT&T and.tl;/
ine being developed by Satellite Business Systems,
dominate the network market and thereby create the kind of
coherence that IBM has created/n a large part of the computer
software field.
. However, the other factors seem to work against coherent
/nterconnection. F/rst, the network situation is evolving with-
out any national policy.
evolvg thout any federal poller.) Sveral counties are
building networks independenfiy. 'Many compardes are
building networks independently. Second, although the lines
may .e leased from the same tlephone company that leases
lines to everyone else, there is some security in having one's
own dedicated network. The' need for security in such areas
as EFT may be stronger than the ned for interconnection.
Third, wherever personal informaon is concerned, and
especially in the federal government, privacy has become a
major issue, and a simplistic interpretation of the privacy
problem sets interconnection into opposition with privacy.
Fourth, research and development in the area of network
security and privacy assurance have not been and are not being
supported at a high enough level to create-soon enough-a
technology that w/1i let one say: "You can have both intercon-
nection and security, both interconnection and privacy; you
can have your cake and eat it, too." Actually, the technology
of communications security is rather well developed [41]-
[44] -except for uncertainties arising from the "56-bit contro-
versy" [45] -as a result of many years of work in the military
intelligence area, but the technology of computer security is
less well developed [46], and nontechnological aspects-plant,
personnel, and operational aspects-of network security are
not in good condition at all.
The conclusion with respect to unification, federation, or
fragmentation must be: we should strive for the kind of feder-
ation of networks that will provide coherent interconnection
where needed and justified and, at the same time, provide
informational privacy and security. That will require planning
at national and international levels. It will require intensified
research and development in network security and in inter-
netting. And it will require an elevation of the ongoing dis-
course about privacy-to a level on which legislative and ad-
mlnl.qtrative policies can be defined clearly ad networks can
be designed responsively.
D. Privacy
It is now very ,ridely understood that the collection of large
amounts of personal information in computer processible data
banks tends to $eopard/ze personal privacy. The main reason,
of course, is that aggregation and computerization open up the
possibility of invading privacy efficiently and on a massive
scale. At the same time, they open up the possibility of pro-
teeting privacy tirelessly and algorithmically and of using the
personal data effectively in the effort to accomplish the legiti-
mate purposes for which the data were collected. Indeed, th&
stage is set for a battle between the forces of good and evil.
The stage is, however, not set in reasonable balance. The
things required for the protection of privacy in information
net. works are policy and technology: legislative and admin-
istrative policy to define what is to be achieved and a tech-
nological basis for achieving it. In the U.S., the legislation is
the Privacy Act of 1974 and the administrative policy is an
OMB Circular [47]. Both are cast in terms of absolutes. The
technological basis, as mentioned in the preceding section, is
a combination of communications security and computer
security. Because computer security is a relatively new and
neglected subject, it is difficult to prov/de convincing assur-
ance to an intelligent skeptic that any proposed intercon-
nection of personal data, transmission channels, information
processors, and interrogation-and-display facilities will not
jeopardize privacy. Repeatedly, indeed, the advocates of pro-
posed federal data networks have failed to present convincing
analyses of the threats to privacy and of the trade-offs be-
tween privacy and mission effectiveness-and, repeatedly, their
requests for permission to procure such networks (e.g., FED-
NET, NCIC upgrade, IRS Tax Administrative System) have
been sidetracked or dnied with good reason.
At least in government circles, therefore, the issue of privacy
is a very real and central network issue. Before it can be
solved, three th/ngs have to be done. l) The technology of
information security has to be improved to the point at which
reasonable analyses can be made and assurances can be given.
2) Network advocates have' to develop plans and justifications
that take privacy into account and provide strong assurances
that it w/li be protected. And 3) the members of the oversight
committees and their staffs have to face the fact that to pro-
tect privacy by precluding interconnection is not a very satis-
factory solution for the long term. They should support and
foster the accomplishment of steps l) and 2) with the aim of.
receiving plans and proposals that they would not have to kill.
E. Other Issues
Space limitation precludes substantial discussion of other
issues, but there are several that should be discussed. We shall
discuss only a few of them, and those only very briefly.
1} Transborder Data Flow: Several European countries are
beginning to restrict the flow of personal (or personnel) data
across their borders on the ground that they must protect the
informational privacy of their citizens against threats implicit
in data processing in countries with leas stringent privacy laws.
Some of the countries have laws or regulations that preclude
the transmission of encrypted information through their
public communication facilities. Many believe that such re-
stfictions may be used to discriminate against foreign (e.g.,
------------------------------<page break>-----------------------------
1344
, American) data proceasing firms and against inuliinational
corporations.'
2} Technology Export and Import: International networks
can be expected to facilitate greatly the transfer of scientific
and tclmical information and know-how among the tech-
nologically 'developed nations. From a nationalistic point of
view, one can see both advantages and disadvantages in such
transfer to ideological, military, and economic competi.tors.
The advantages are mainly humanistic and short-term eco-
nomic. The disadvantages are mainly security-related and
long-term economic. The'interplay of advantages and dis-
advantages is giving rise to issues that will probably intensify.
3) Compea'tion Versus Monopoly and Free Enterprise
Versus Regulation: These are the issues of the "Bell Bill"
(Consumer Communications Reform Act), Computer Inquiry
II, and several recent decisions of the Federal Communications
., Commission that have favored compet2tion in the telecom-
munications industry. How these issues are settled will to a
large extent determine who operates the networks of the
future in the U.S. and how such applications as electronic
message service and "te office of the future" ar imple-
mented. [
4; Nature of Office Pdork and Pdorkforce: Some of the net-
work applications we have discussed would tend to. alter
markedly the nature of white-coliar work and the knowledge
and skills required of members of the workforce.. That fact
will give rise to issues involving reeducation and retraining, pay
commensurate with responsibility, and displacement of labor
by automation.
$ Impact on Produca'vity: Many people are expecting that
applications such as electronic mail and office automation will
ß significantly increase productivi.'ty, but there are as yet few if
ß - any definitive experiences with such applicatibns or quantita-
ß tive models of them that will convince skeptics. Impact-on-
productivity may become a major issue in and of itself.
6 Educatl'onal Applications of Networks: Packet-switched
and satellite networks, togetherwith the great advances being
made in computers, appear to open the door to revolutionary
improvements in education, but much more than mere access
and mere hardware will be required to achieve truly significant
results. The issue that is arising is whether the society values
education enough to support the long and difficult effort that
will be required to develop effective computer- and network-
based methods-or whether there will be' another wave of pre-
mature exploitation followed by disappointment as there was
in the computer-assisted-instruction "revolution" of the early
1960's.
ß .i:...2:.i V Networks 'Versus Stand-Alone Systems: Why do we need
[:._-time sharing when everyone can have his or her own micro-
'0mputer?' What good is a network when one can have a
.111}h01e library on a video disk? Those questions have answers,
ß : Of course, in such applications as electronic message systems,
ß '- distributed but cooperating "offices of the future," and
Computerized commerce, but the questions will nevertheless
.constitute a major issue. Microcomputers and inexpensive
-. digital storage devices have significantly changed the network
concept. Les than a decade ago, a computer network was
something that provided access to a time-sharing system. Now
it is a facility to support communication among spatially dis-
tributed .people and computers and to supply people and
computers with common information bases and supple-
.storage and processing capabilities.
An important latent issue is implicit
people have quite different percep-
PROCEEDINGS OF THE IEEE, VOL. 66,'NO. 11, NOVEMBER 1978
tions of the importance of networking. A significant fraction
of the people who have had experience as developers or inten-
sive users of a packet-switching network believe they have
been in on the beginning of a new era and that descendants of
the ARPANET will constitute the nervous system of the
world. On the other hand, most of the people who now deter-
mine the kind of national policy that earlier fostered the
merchant marine, the railroads, the airlines, and the interstates
seem not to be aware that any significant new potential exists
or that there may be any reason to move rapidly to take
advantage of it. -And, of course, if it is meanilagful at all to the
man in the street, the term "information network" still
suggests the telephone system, the radio, or'a television net-
work.. In that situation, it is difficult to project as an issud the
importance of networks to world economic leadership. We
believe, nevertheless, that it is such an issue, and we hope that
it will soon be recognized as such an issue.
9) Totalitarian Control: If almost all the telecommunica-
tions in an area were based on computerized networks con-
trolled by an organization--say by a government-then, in the
absence of effective safeguards, that organization could map
the life space of every individual and record the business trans-
actions of every company. The notion of telecommunications
in the hands of a "big brother with computers" goes beyond
the bounds of what is usually called "invasion of privacy" into
the realm of totalitarian control. One can detect at least a
trace of the "big brother"issue in the coldness of certain
members of Congress toward, and the rejection by the OMB
of, the plans (mentioned earlier) of the IRS to develop a
computer- and network-based Tax Administration System.
The mere possibility of subversion was enough to kill the
system. It is of the utmost importance, of course, to develop
truly effective safeguards against misuse of networks for pur-
poses of soc/al control. But such safeguards will be more
difficult to devise than safeguards against ordinary invasion of
privacy or against fraud a/nnd theft. Networks will have to be
designed in such a way that representatives of d/verse interests
can satisfy themselves that there is no sabversion and that the
audit trails are not dosslets. And the arrangements will have to
be dictator-proof. We think that that is a very great task and
that it is being neglected.
V. CONCLUSIONS
Shakespeare could have been foreseeing the present situation
in information networking when he said," . . . What's past is
prologue; what to come, in yours and my discharge" [48].
Most of the applications that will shape the future of net-
working are now in the stage of conceptualization or in the
stage of early development. But it seems possible that a "net-
work of networks" will, even in this century, become the
nervous system of the world and that its applications will
significantly change the way we live and workß The degree to
which the potentials of networking w/11 be realized will depend
upon how we resolve some of the issues that have been
discussed.
The value of information networks will depend critically
upon their connectivity and their ability to connect any one
of many sources to any one or more of many destinations.
High connectivity will be precluded if conditions force the
development of many separate, independent, incompatible
networks. One condition that would force such an incoherent
development is the combination of. 1) a need for security
against loss of "electronic funds" and (other) proprietary
------------------------------<page break>-----------------------------
AND VEZZA: APPLICATIONS OF INFORMATION NETWORKS 13,
and 2) the lack of a technology capable of pro-
security in an interconnected network or network.of
That would lead to what we have called "frag-
Another such condition is based in a .imilnr way
combination of need for informational privacy and lack
ithe technology necessary to protect it except by isolating
,-pivacy-sensitive data. The rapid and intensive develop-
of computer and network security technology is vital to
network applications.
forces are fostering the develc;pment of networks to
organization o r branches of organizations and the
of applications to serv oanitions, but there
few forces that foster networks to interconnect individuals
inetwork applications to serve individuah. Perhaps the main
)e for the provision of network services to individuals-
ecially network services to individuals at home-is that the
System will move (as obviously it would like to do) into
procsing, storage, and information-commodity parts of
overall information business. But the telephone companies
be very slow to provide high-information-rate services
they have such large investments in narrow-band
To get inexpensive wide-band channels into homes
early date, we need a new departure in cable (or fibe
communication, taking off from cable television, dr
truly revolutionary like a nation-wide network 'of
platforms: microwave platforms at 70 000 ft,
orted by helium plus helicopter vanes, and relaying signals
housetop "dishes" a meter in diameter.
Examination of 30 actual and potential applications of net-
suggests that the following network characteristics or
are especially important: bidirectionality, freedom
ted errors, efficiency despite "burstiness" in the
pattern, inherently low cost, high connectivity,
transmission rate, security, privacy, cuthen-
and reliability. Mobility and broadcast capability
out to be of the lowest priority in our analysis. Packet-
and time-division-multiple-access networks, espe-
such networks with satellite relays, were suggested by
analysks to have the patterns of characteristics required to
best the full range of applications. The analysis suggests
roach to the selection of the best network to serve any
or set of applications.
REFERENCES
i] T. H. Myer and C. D. Mooera, HERMES User's Guide. Cam-
bridge, MA: Bolt Bernek and Newman, Inc,June 1976.
A. Vezza and M. S. Broos, "An electronic message system: Where
does it fit?" in Proc. IEEE Conf. Tends and Applications 1976:
ComputerNetworks, New York, Nov. 1976.
A. Vezza, "A model for an electronic postal system," in Telecom.
i' municatlons Policy Researel Conf. Proe., Bruce M. Owen, Ed.
 Palo Alto, CA: Aspen Institute Program on Communications and
Society, Apr. 1975, ppl 1.6-153. -
SIGMA Mes. tage Service: Reference Manual, Information Sciences
Inst., Univ. Southern California, Marina Del Ray, CA, ISIfrM-
78-11.
;] J. B. Rule, "Value choices in electronic funds transfer policy,"
 Office of Telecommunications Policy, Executive Office of the
President, Washington, DC, Rep. Domestic Council Committee
on the Right of Privacy, Oct. 1975.
] T. M. Dunn and J. H. Mirrissey, "Remote computing-An experi-
mental system. Part 1: External specification.a," in Proc., Spring
Joint Comlurer Conf.. 1974.
J. M. Keller, E. C. Strum, and G. H. Yang, "Remote computing-
An experimental system. Part 2: Internal design," in Proc Spring
Joint Computer Conf., 197..
J. E. Sammet, Programming Languages: History and Funda.
mentals. Englewood Cliffs, NJ: Prentice-Hall, 1969, pp. 226-229.
T. Marill and D. Stern, "The datacomputer--A network data
utility," in AFI_P$ Conf. Proc., Montvale, N J: AFIPS Press, May
19'5.
[10] L. G. Roberts and B. D. Wessler, "Comput'er network develop.
merit to achiev resource sharing," AFIP$ Conf. Proc., May 1970.
[I 1] L. G. Roberts, "ARPA network implications," ARPA Network
Information Center Doe., 12 982, 19ff2.
[12] R. M. Metenile, "Packet communication," M.I.T. Ib. for Com-
puter Science, MAC/TR-114, Dec. 1973.
[13] "SNDMSG,"Tenex User's Guide, Bolt Betarick nd Newman,
Inc Cambridge, MA, Oct. 1977.
[14] J. Vallee, "The forum project: Network conferencing and its
future applications," Cornput. Networks (Nerheriand), vol. 1,
no. I, pp. 39=52, June 1976.
[15. M. Turoff, "Party-line and discussion: Computerized confererlce
systems," Office of Emergency Preparedness, Washington, DC,
Rep., Apr. 1972.
[16]  "Delphi conferencing. Computer-based conferencing with
anonymity," in Technological Forcagting and Social Change.
Washington, DC: National Resource npanalysis Center, 1972, vol.
3, no. 2, pp. 159-204.
[17] M. Turoff and J. Scher, "Computerized conferencing and its
impact'on engineering management," in Proc. 23rd Annu. Engi.
neering Management Conf. on Effective Management of Engi-
neering Resources, Washington, DC, Oct. 1975.
[18] D.C. Engelbart, "Coordinated information services for a dis-
cipilne- or mission-oriented community," in Computer Communi-
cate'on Networks, R. L. Grimedale and F. F. Kuo, Eds. Th6
Netherlands: Nootdhoff International, 1975, pp. 89-99.
[19] D.C. Engelbart, R. W. Watson, and J. C. Norton, "The aug-
mented knowledge workshop," in AFIP$ Conf. Proc., vol.
New York, June 1973.
[20] --, Advanced Intellect. Augmentation Techniques. Menlo Park,
CA: Stanford Research Institute, Feb. 1972.
[21] --, "Human intellect augmentation techniques," Stanford Re-
search Institute, Menlo Park, CA, Final Rep., 1969.
[22] R. Flesch, The Art of Plain Talk. New York: Harper & Blothem,
19.6.
[23] J. O. Mason, Jr., "Management-information--systems-the audi-
lot's role," TheInternal Auditor, vol. 32, no. 5, pp. 40-48, Sept./
Oct. 1975.
[2t] S.J. Pokeroper, "Management information-systems--A pragmatic
survey," Conf. Board Rec., vol. 10, no. 5, pp. 49-54, May 1973.
[25] R. G. Schroeder, "A survey of management science in university
operations," Manage. $cœ-Appl., vol. 19, no. 8, pp. 895-906,
Apr. 1973.
[26] P. A. Stressman, "Managing the costs of information," Harvard
Bus. Rev., vol. 54, no. 5, pp. 133-142, Sept.-Oct. 1976.
[27] G.H. Barnes, R. M. Brown, M. Karo, D. J. Kuck, D. L. Slotnick,
and R. A. Stoker, "The l]liac IV computer," IEEE Tran&
Cornput., vol. C-17, pp. 746-757, 1968.
[28] W. J. Bouknight, S. A. Denenberg, D. E. Mcintyre, J. M. Randall,
A. H. Sameh, and D. L. Slotnick, "The Illiac IV system," Proc.
IEEE, vol. 60, pp. 369-388, 19'72.
[29] "elf-teaching design and analysis tools offered on timeshared
basis," Cornput. Des., vol. 17, no. 2, pp. 18-20, Feb. 1978.
[30] I. E. Sutherland, Computerized Commerce. Santa Monica, CA:
The Rand _Paper Series, Rand Corp., Sept. 1975.
[31 ] National Research Council, Telecommunication for metropolitan
areas: Near-term needs and opportunities. Washington, DC: Nat.
Acad. Sci., 1977.
[32] L. L. Weed, "Technology is a link, not a barrier, for doctor and
patient," Modern Hospital, vol. 114, pp. 80-83, Feb.'1970.
[ 33 ] --, Medical Records, Medical Education, and Patient Care. The
Problem-Oriented record as a Basic Tool. Chicago, IL: Year
Book Medical Publishers, Inc,1970.
[34] R. Davis, B. Buchanan, and E. Shortliffe, "Production rules as a
representation for a knowledge-based consultation program,"
Stanford Univ. Artificial Intelligence Lab., Stanford, CA, Memo
AIM-266, Oct. 1975.
[35] H. Silverman, "A digitalis therapy advisor," M.I.T. Lab. for
Computer Science, MAC/TR-lt3, Jan. 1975.
[36] "GECOS: General Comprehensive Operating Supervisor," Honey-
well Information Systems, Waltham, MA, Rep. D D 19.
[37] A. Stathopoulos and H. F. Caley, "The Autodin rl network,"
in EA$CON-77 Rec. Proe. IEEE Electronics and Aerospace Sys-
tems Convention (Arlington, VA), Sept. 1977.
[38]' Social Security Administration, Office of Advanced Systems,
Recommended Design Concept for the Future S$A Process,
Washington, DC, OAS Pub. No. 005, Apr. 1977.
[ 39] V. Slamenka, private communication, May 1978.
[40] CCITT, "Recommendation X.25: Interface between data terml-
hal equipment (DTE) and data circuit-terminating equipment
(DCE) for terminais operating in the packet mode on public data
networks," in Public Data Networks, Orange Book (vol. VII.2).
Geneva, Switzerland: Sixth Plenty Assembly, International Tele-
communications Union, 1977, pp. 70-108.
[41] W. Diffie and M. Hellman, "New dh-ections in cryptography,"
IEEE 7'rans. Inform. Theory, pp. 644-654, Nov. 1976.
------------------------------<page break>-----------------------------
1346
[42] R. Rivest, A. Sinamir, and L. Adleman, A method 'for obtaining
digital signatures and ImbHc-key cryptos)stems," M-I.T. Lab. for
Computer Science, LCS/TM-82, Apr. 1977.
[43] FederalRegister, vol. 40, no. 52, Mar. 17, 1975.
44] FederslRegi. rter, vol. 40, no. 149,Aug. 1,1975. '
45] W. Dtffle and M. HeUman, "Exhaustive cryptanalysis of the NBS
data encryption standard," Computer, vol. 10, no. 6, pp. 7a-sa,
PROCEEDINGS OF THE IEEE, VOL. 66, NO. 11, NOVEMBER 1978
June 1977.
[46] $. H. Saltzer and M.D. Schroeder, "The protection of informa-
tion in computer systems," Proc. IEE, vol. 63, pp. 1278-1308,
Sept. 1975.
147] Office of Management and the Budget, OMB Circular A-108,
Washington, DC, July 1975.
[48] W'filiam Shakespeare, The Tempext, act II, scene I.
A Tutorial on Protocols
!
LOUIS POUZIN D HUBERT ZIMMERMANN
Invited Paper
Abract-Protocols are common tOOls for controlling information
transfer between computer Systems. The concept of a pintocot, which
gew out of experimental computer networking, is now fundamental to
system design. In this paper, basic protocol functions ate explained and
discussed. Then, the concept of a diatn'buted system architecture is
presented. It providea the framework for layers and protocols to operate
across heterogeneous Systems. The purpose and functions of each
protocol layer $uch aa, tranmliaaion, transport virtual temtinal, axe
descan'bed. Interactions between design and performance axe discussed,
and typical mechanisms are zriewed. CCITT and ISO relevant standaxcks
are summarized. Finally, the similarity between protocol$ and program
ming languages is emphasized as it points to the major impact brought
about by protocol$ in system design.
L INTRODUCTION
VERYONE has had the opportunity to overhear such
cryptic conversations exchanged over the radio by taxi
drivers, policemen, and aircraft pilots. Although upon
hearing these conversations at fn-st do not mean much to the
layman, these abbreviated languages carry well-defined mean-
ings and obey well-der'meal rules. Speakers give their name, ask
correspondents if they are listening', confirm reception, etc.
This form of conversation differs drastically from a face-to-face
chat. The communication channel is shared by many speakers.
To save bandwidth and reduce haterferences, messages are short
and coded. External noise and other interferences are com-
mon occurences, hence, repetition and confzrmation are normal
practice. These rules are known as protocols.
Manuscript received February 27, 19'78; revised July 114, 1978.
The authors are with IRIA, 78150 Rocquencourt, France.
The term protocol entered the computer jargon at the turn
of the 70's, when the U.S. Defense Advanced Research Project
Agency set out to build a network of geograpkically distribute
heterogeneous computers [71]. Up to that time, communica-
tion between computer programs or processes was limitec to
processes which were located within the same machine. Inter-
process communication was accomplished through the use of
shared memory and special signals exchanged through the
mediation of the operating system. Titis technique represented
the analog of a face-to-face chat between processes. Inter-
process communication between geographically distant systems
would have left processes with the same kind of constraints
that taxi drivers encounter. They would have to interact
through a potentially hostile environment with limited band-
width, delay and unreliable transmission. In addition, the
processes in the different computer systems did not even speak
the same native tongue, having been created by different
manufacturers.
Computer veterans remember the sinuous evolution that led
from binary programming to assembly code, to Fortran', Cobol,
Algol, and other high-level languages. Originally viewed as a
collection of tricks and hobbies, programming languages have
developed into a major branch of computer science. The
evolution of protocols has followed a strikingly similar path.
Indeed, Protocols are common tools designed for controlling
information transfer between computer systems. They are
made up of sequences of messages with specific formats and
meanings. These messages are equivalent to the instructions
of a programming language, although protocol languages are
still in an early stage of ad hoc development.
0018-9219/78/1100-1346500.75 ¸ 1978 IEEE
------------------------------<page break>-----------------------------
------------------------------<page break>-----------------------------