The outlook for computer mail

The outlook for computer mail

Raymond R. Panko

The a u t h o r aims to present a balanced view of the prospects for computer mai l , and to br ing ou t the i m p o r t a n t po l i cy issues su r round ing the new medium. After descr ib ing, w i t h the aid of a scenario, typical app l i ca t i ons of t he sys tem, he ou t l i nes h is tor ica l developments and cu r ren t t rends. He then discusses potential market size and develops cost p ro jec t ions . The au tho r sees lack of c u s t o m e r awareness as a serious obstacle to a w i d e r use of computer mai l and b a l a n c e s t h i s a g a i n s t a cons ide ra t i on of the relation of t he sys tem to o rgan i sa t i ona l needs. F inal ly , he ana lyses the policy c o n c e r n s r e s u l t i n g f r o m t h e development of computer mail.

The author is wi th the Stanford Research Institute, Men lo Park, CA 94025 , USA, and the Universi ty of Hawai i , Honolulu, Hawai i 96822 , USA.

The research described in this article was supported in part by Bell Canada.

1John Holusha, 'Computer tied Carter, Mondale campaigns', The Washington Star, 21 November 1976, pA-3. 2 Raymond R. Panko, 'The outlook for computer message services: A preliminary assessment', unpublished draft, Telecommunications Sciences Center, Stanford Research Institute, Menlo Park, California, March 1976.

With the advent of time-sharing, computer users suddenly found themselves hundreds of yards, or even miles, from operators and other users. System designers responded to resulting communication needs with 'mailbox' programs, which allowed users to exchange brief messages. Later, the growth of national computer networks created even greater communication stresses, and the clumsy mailbox services then in use grew increasingly inadequate. Commercial network operators developed some new message services, but the first real surge of progress came in the 1970s, when the US Defense Department linked major US research computers through the A R P A N E T computer network. Computer scientists soon adapted their local mailbox tools to network delivery and began making improvements. In this open and well funded environment, mailbox systems evolved with amazing speed. Borrowing software from office automation systems, A R P A N E T mail designers soon had simple yet powerful systems that automated the full life cycle of a message, from initial composition and editing through reading, filing and retrieving.

These new systems - which are called 'computer mail' to distinguish them from their mailbox predecessors - soon attracted people who had never worked with computers before. Small organisations began adopting computer mail for mainline office communication and soon, some network observers claim, message traffic dominated the ARPANET ' s use.

Today, a litany of computer mail users (among them President Carter of the USA ~) would be brief but impressive. But it is the future that catches the eye of everyone who looks at computer mail. The current cost of an average message is reasonably high, comparable to the cost of Telex or TWX, but costs are falling rapidly, and computer mail should be cost-competitive with postage by the mid 1980s. 2 Offering instant delivery, simple operation, and life-cycle automation, computer mail has a good chance to compete with all forms of written communication. It is estimated that well over 500 billion pieces of written communication flow in the USA each year, and the actual number may be much larger.

Despite this potential, and despite the advanced state of existing software, computer mail remains an invisible revolution. The A R P A N E T environment, in which most development has taken place, is a cloister, within which colleagues talk, yet beyond which information flows only in trickles. Moreover, few A R P A N E T mail

242 TELECOMMUNICATIONS POLICY June 1977

Z l should point out that I use computer mail myself, w i thou t going through a secretary. This is nei ther typical nor atypical: use o f compute r mail is highly id iosyncrat ic and it is impossible to tell, on the basis of a person's exper ience wi th computers or their organisat ional level, how a given individual wi l l adjust to computer mail. 4This is a two - or th ree-minute ordeal, in which I dial the computer network, tell it wha t kind of terminal I have, tell it wha t computer I want , tell the computer who I am, and ' te l l the computer that I wan t to read my mail. The process wi l l undoubted ly be simpl i f ied in t ime, but it is current ly a major imped iment to system use. s The calendar system sends messages all the t ime, the moni to r in forms me now and then that one of my old f i les has been archived on to tape, and so on.


designers have been concerned about their medium's market potential. This is beginning to change, as symbolised by the recent commercial offering of one A R P A N E T mail system, HERMES. But it is probable that computer mail will be economically attractive long before its commercial potential is actually exploited. If that is true, computer mail's ultimate entry may be quite explosive.

So far, I have talked only of reasons for optimism, stressing the availability of sophisticated software, a favourable price outlook, the existence of a huge potential market, and computer mail's good competitive position relative to postal delivery. But there are dark clouds on the horizon too, including a murky regulatory outlook, confusion about future directions among system designers, and a debilitating absence of consumer awareness.

Typical applications A good way to describe computer mail is to give a scenario for its use. I will try to do this with a hypothetical session that exemplifies my personal experiences with computer mail) The following scenario concerns a person using the system directly, first, because this is probably the most cost-effective way to use it in an informal environment; second, because this illustrates the full power of the medium; and third, because that is how I use it.

The hardest part of using computer mail is the still-clumsy process of 'logging in' to the computer. 4 Once the computer understands me, it tells me if I have new mail. If I do, it gives me a one or two line summary of each message, including author, date, and subject. I usually receive about two messages per day, but to illustrate the process, I will assume I have received five.

The first message is electronic junk mail: a weekly progress report for a project loosely related to mine. It is sent to me as a courtesy, and I normally delete it immediately. But every now and then, like today, I want to scan it. Since it is three pages long, I tell the system to list it. It appears a few minutes later on a high-speed printer, and I will scan it at my leisure.

The second message is from Ted Myer, a mail system designer at Bolt, Beranek and Newman, in Cambridge, Massachusetts. He tells me that he will visit Stanford Research Institute next Wednesday and asks if we can meet. I type the 'answer' command, then type a short reply. It is delivered to him immediately. I send a message to my calendar file, noting that I will be meeting with him, and I tell the calendar to notify me the day before he arrives.

My third message is from the calendar, reminding me that today is my wife's birthday, that I have reservations for dinner, and that I should call the florist. Many uses of computer mail are as informal and personal as this one. Also, the fact that the computer sends me a message is not unusual. 5

The fourth message is from Jim Carlisle at the University of Southern California. Jim has broadcast this message to fifty people around the USA, telling them that a new paper on computer mail is available. If I want it immediately, I must go through a complicated process to ship it across the network to my computer. Sadly, few systems can handle long documents effectively. Because the paper is not urgent, I send a brief reply, asking him to mail a copy. Then, with a single command, I foreward a copy of the message to a third party, who

TELECOMMUNICATIONS POLICY June 1977 243


6Western Union, letter to the Federal Communicat ions Commission, 14 March 1966, Re: Computer Leases and Service Arrangements, p 12, quoted in Stuart L. Mathison and Philip M. Walker, Computers and Telecommunications: Issues in Public Policy, Prentice-Hall, Englewood Cliffs, N J, 1970.

might be interested in the paper. The fifth and last message responds to an inquiry I broadcast last

week to several colleagues on the A R P A N E T , asking them when various mail systems became operational and who developed them. The sender of this message is someone I have never met before. From the A R P A N E T directory, I find that he is with the University of Wisconsin. After sending a brief on-line reply, I jot down his telephone number and plan to telephone him tomorrow. Like several other people I have met through computer mail, he may become a close colleague in the future.

I have now finished responding to the messages sent to me today. I also wish to send a message to Larry Day, of Bell Canada. I give the command, 'send' and the computer asks me to whom the message will be sent. I type the name 'Day' , but because there are many people with that name on the system, the computer types out a short list of them, giving their full names, their organisational affiliations, and identifiers that function much like telephone numbers. I type in Larry's identifier. The system asks me if I want copies sent to anyone, and asks me to supply a brief title. It then lets me type and edit the message. (Alternatively, I could specify a pretyped document for transmission.) The process is very much like filling out a memorandum. I reread the message, than transmit it. Because it is not urgent, I queue it for overnight delivery.

It occurs to me that Jim Carlisle of USC asked a question last week and that I could not reply immediately. I have the system scan through all messages from Carlisle. The system prints a header for each Carlisle message, including the date and subject. I stop the printing when I see the message I want, then type it out and reply to it. I also tell the system to send a copy of the message to another friend, who is not a computer mail user. The system prints a copy of the message on the high-speed printer, together with a mailing label. My secretary puts it in the afternoon mail.

My session is now over. I have spent 12 minutes on-line, and I have sent two messages. The average message was about 50 words. Counting copies sent to people other than the primary receiver, I sent each message to three people. This is reasonably representative of the experiences of most users.

Historical developments The use of computers in public message services began in the 1960s, when several Telex and T W X suppliers started to automate their switching functions. Private users were also quick to grasp the advantages of computerised message routing. As early as 1965, Western Union could identify 57 organisations with private, computer-based message switching systems. 6 Today there are many computer-based teletypewriter systems, including large military systems, systems for individual companies, and some industry-based systems.

On the whole, these systems view the computer only as a sophisticated switch. None, to my knowledge, offers interesting composition or reading tools. Only a handful allow users to search through old messages. Yet the very fact that many organisations and public carriers already have computer-based teletypewriter exchange networks means that a potentially, large user base for advanced


Based on several speeches by Russell W. McFall, president of Western Union, mentioned on p 20 in Mathison and Walker, op cit, Reference 6.

Mathison and Walker, op cit, Reference 6. t The Compatible T[me-sharing System. A Programmer's Guide, 2nd edition, edited by P.A. Crisman, Massachusetts Institute of Technology Press, Cambridge, Mass, 1965, Section AH.9.05, quoted in Mathison and Walker, op cit, Reference 6. Io Most developments were bootlegged on other projects, but several, for example the NLS Journal Mail system at Stanford Research Institute, were directly funded by ARPA.


computer mail systems may be readily available without drastic disruptions in corporate communication purchasing practices.

In the middle and late 1960s, several communication carriers became very interested in selling data processing along with their communication services. In the USA, Western Union publicly staked its future on becoming an information company. ~ At the same time, several data processing companies began to put toes into communication waters. Bunker-Kamo wanted to offer order-handling services, along with its information services, for the financial securities community) But, for a large number of political and pragmatic reasons, data processing and communications have generally remained apart throughout the world. In addition, most companies simply wanted to offer both message switching and data processing; none seriously explored the use of the computer to augment message preparation or disposition.

The most steady stream of development in the use of computers for communication has come from computer scientists. The development of remote, interactive computing in the mid 1960s meant that users were often far from operators and other users when problems arose. Consequently, most system builders developed 'linking' tools, which essentially let two users lock their terminals together so that each could see what the other was typing, and 'mailbox' tools, which allowed brief messages to be sent to on-line users or to the working spaces of users not on-line when the message was sent. The very first interactive time-sharing system, CTSS at MIT's Project MAC, had both these tools, 9 and it is rare to find any time-sharing computers without them.

In the late 1960s, the US Department of Defense began to link contractors of the Advanced Research Projects Agency (ARPA) with a computer network called the ARPANET. Network mail delivery began almost immediately, and by 1975 a large number of rather sophisticated mail services had been built. ~°

By 1974, ARPA considered the services sufficiently advanced to test them in an operational se t t ing outside the ARPANET community. ARPA and the US Navy agreed to perform a joint experiment in Hawaii, and ARPA funded the final development of three mail programs, one at MIT, another at Bolt, Beranek and Newman, and the third at USC's Information Sciences Institute. These three systems have created state-of-the-art standards for mail software, and it seems very likely that ARPA funding of computer mail will continue in the future.

The late 1960s brought another development stream, computer teleconferencing. Originally, the Delphi technique, pioneered in the 1960s for the pooling of expert opinion, was the model for teleconferencing. It was envisaged that Delphi questionnaire- answering systems would be constructed to pool experts in different locations. But when the first of these systems was developed, by Turoff at the US Office of Emergency Preparedness, it became clear that the experts wanted to talk directly with one another.

The computer quickly became the group's scribe, passing around private messages and posting public messages in an on-line record. Participants came and went as they pleased, consulting the on-line record to keep abreast of the dialogue, and typing in their own comments as appropriate.

By 1973, the FORUM conferencing system, developed at the



Institute for the Future (largely under ARPA and National Science Foundation funding) was the best system. Between 1973 and 1975, FORUM was tested in almost 30 conferences. These experiments were studied under NSF grants, and improvements were rapidly incorporated. In late 1974, a variant of FORUM called 'Planet- 1' was offered, on a restricted basis, over TYMNET. At about the same time, Bert Liffman's General Conferencing System (GCS) was offered on the I.P. Sharpe network in Canada. In 1976, the Electronic Information Exchange teleconferencing system was developed under Turoff at the New Jersey Institute of Technology. It is now being used for experimental conferences via Telenet.

Computer conferencing has received more publicity than computer mail, largely because its development was guided by publishing scientists rather than artisans, yet it is certain that computer mail has handled several orders of magnitude more traffic than computer conferencing.

1tin 1974, for example, the bulk of the quarter billion messages sent on the international airlines network, SITA, were administrative messages. See Phil Hirsch, 'SITA: rating a packet-switched network', Datamation, Vol 20, No 3, March 1974, pp 60-63. lz Mai lgram in the USA and Telepost in Canada, which link Telex and TWX to postal delivery, have indicated how computer message services can be used to reach people who do not use the message service.

Trends The future development of research systems is impossible to predict, in part because traditional design paradigms and distinctions are breaking down. When computer mail and conferencing systems are employed in real organisations for prolonged periods, for instance, their basic statistics, including average message length and the ratio of group messages to messages for individuals, tend to be quite similar. Moreover, as discussed below, it is not clear that either the teleconferencing model or the computer mail model, as they have traditionally been conceptualised, will lead to computer-based communication systems capable of serving the real communication needs of organisations or of exploiting the potential for computers in communication.

Most commercial time-sharing networks offer relatively crude mailbox programs, which do not rival ARPANET mail software. But two commercial networks have already adopted conferencing or computer mail software. Furthermore, some commercial networks, as in the case of Telenet, are linking users to host computers offering ARPANET- or NSF-developed mail and conferencing software. TYMNET is taking another route, designing its own advanced mail program, and other networks may follow suit. Xerox and IBM are also designing computer mail Systems, but these are likely to be next- generation services and may not reach the market until after 1980.

Other trends appear to be converging. Computer systems for management information, ordering, accounting, inventory control and similar purposes have often been used in part for message communication, usually to the confusion and dismay of their data- centred designers. I~ Many of these data-base systems may adopt computer mail. In another trend, many new word-processing typewriters are being sold with communication hardware that could be used to interface them with a computer mail network. ~2 Finally, both NLS and Graphnet (a US value-added network) allow graphics as well as text to be handled.

Many trends may be converging, and these could spur the evolution of computer mail. But, to keep things in perspective, it must be borne in mind that all the prospects are highly speculative. Business and government users have yet to probe the adequacy of

246 T E L E C O M M U N I C A T I O N S POLICY June 1977

~a It should be noted that available data are incomplete and possibly inconsistent. ~4 President's Commission on Postal Organizat ions, Towards Pos ta l Excellence, US Government Printing Office, June 1968. ~5 J.R. Hinrichs, 'Communications activity of industrial research personnel', Personnel Psychology, Vol 17, November 1963, pp 194-204. See also 'V~/P: path to the 1980s', Modern Office Procedures, Vo120, No 6, June 1975. is 'Better procedures improve paperwork', The Office, Vol 77, No 5, May 1973, pp 76-77. 1~ John P. Scheid, 'A brief history of the forms industry', The Office, Vol 77, No 5, May t973, pp 57-59, and 102. 1e John W, Haslett, 'The new trend is people paperwork', The Office, Vo177, No 1, January 1973, pp 71-72. See also Joseph Ferriera, and John Marus, "Regulation: more heat under business, Computer Decisions, Vol 8, No B, August 1976, pp 20-29. 19 Personal communication with Robert Grabau, Manager, Marketing Research, Moore Business Forms Corporation. Z°Daniel P. Lavery, 'Trends in copier usage', The Office, October 1974, pp 81- 83. See also Ken Ford, 3"he development of our office copying process', The Office, Vo177, No 6, June 1973, pp 18-20. zl Ford, op cit, Reference 20.


current design notions, and short-term developments are simply impossible to predict.

Potential market size

There are comparatively few data on the flows of paper in the market economy. However, it is possible to provide a rough estimate of the market potential for computer mail in relation to the country for which most data are available - the USA. ~3

Conceptually, we view an individual worker's world as a series of concentric circles. In the middle is the individual worker in his or her own office, including the person's own head, filing cabinets, desk calendar, notes, and other information stores. Next came the office, the company, and the general organisational community. Even this coarse breakdown is difficult to analyse.

In 1968, Arthur D. Little, Inc studied post office flows for the President's Commission on Postal Organizations. ~4 If that study's percentage breakdowns hold good today, the annual mail flowing among business and government organisations comprises 4 billion pieces of correspondence, 4 billion "advertisements, and 17 billion pieces of transaction mail (of which 3 billion are money transfers).

It appears that interoffice mail within companies is much more extensive than postal mail. On the basis of very limited dat~, interoffice correspondence is two to four times larger than postal correspondence.~5

But interoffice correspondence flows are minor, compared to interoffice flows of forms. Roughly 40% of all office work hours are clerical, and 75% of those are devoted to handling forms. ~6 Form sales alone have grown from $50 million in 1940 to about $2 billion today, and will reach $3 billion by 1980. ~7 Growing government record-keeping requirements are responsible for some of this growth,/8 but general business needs have been responsible for most of it. At an average cost of $5 to $25 per 1000 forms, ]9 between 80 and 400 billion forms now flow through organisations each year, many being transferred several times during their lives.

Because communication systems have not historically supported paper flows within offices, there are virtually no data on intraoffice written communication. The pattern of copier use shown in Table 1, however, may indicate that copier use is dominated by items sent out for general office distribution, for example bulletins and group memoranda. In the USA, 50-100 billion copies are made each year on convenience copiers. 2° In addition, many of the 200 billion pages generated by centralised duplicators each year 21 may also be for intraoffice distribution.

These estimates do not include 'flows' from group filing cabinets to individuals, nor do they include other types of informal intraoffice information flows. In additon, although we tentatively allocated all

Source: Daniel P. Lavery, 'Trends in copier usage', The Office, October 1974, pp 18- 20.

Table 1. Percentage of sessions in which so many copies were made per original.

Office size Number of copies

1 - 5 6-11 12-25 26-100 100+ Large 43% 12% 8% 15% 22% Small 58% 32% 3% 6% 1%



Source: Raymond R. Panko, 'The outlook for computer message services: a prel imi nary assessment', unpublished draft, Telecommunications Sciences Center, Stanford Research Institute, Menlo Park, California, March 1976.

Table 2. Cost projectionsfora bHefmessage.

Item 1975 1977 1985 Method I Method II

($) ($) ($) ($)

Computer 0.30 0-10 0.02 0.05 Computer-centre labour 0.30 0.10 0.05 0,07 Communication 0.30 0-30 0.15 0.03 Terminal 0.30 0.30 0.30 0.10 Total 1.20 0.80 0-52 0.25

22 Nobody, to my knowledge, has studied paper f lows within an individual's personal office space, but it may well be that filing, calendar entries, action notes, and the like are an order of magnitude larger than the f lows of paper into and out of the office. Again based on rudimentary data, it seems that computer mail systems are already being used for i n f o r m a t i o n t rans fe rs w i th in an individual's personal files. z3 Panko, op cit, Reference 2. 24The worldwide, private SITA network, for example, has expenses of only $0 .20 per message. See Phil Hirsch, "SITA: rating a packet-switched network, Datamation, Vo120, No 3, March 1974, pp 60-63. 25 Panko, op cit, Reference 2. 2s Frederick G. Withington, 'Beyond 1984 : a t e c h n o l o g y fo recas t ' , Datamation, Vol 21, No 1, January 1975.

form flows to interoffice mail, it is conceivable that many of these forms really flow within offices. 22

To sum up, it appears that well over 500 billion pieces of paper work flow among and within organisations in the USA each year and, at least in the forms and copier categories, volume seems to be increasing rapidly. Flows are most intense among people who are closest in the organisational structure.

Cost and cost projection

Only one thing is more hazardous than to quote costs, and that is to project them. But we need some insight into current and future costs if we are to speak intelligently about computer mail's potential.

In 1975, the Stanford Research Institute analysed cost data from six message systems. ~3 That study found that the cost per message, for brief messages of around 50 words, was between $0.50 and $1.50. Conventional wisdom holds that this total cost is due, about equally, to computer-centre labour, machinery, and communications. Using empirical connect time data and assuming that users share a teletypewriter terminal used 25% of the day, a terminal charge per message of $0.30 is estimated for this study. For convenience, it is assumed (see Table 2) that labour, machinery, communications, and terminal rental all cost $0.30, for a total cost of $1.20 per message.

This is roughly the cost of Telex or TWX. That a computer-based system should be as economical as Telex or TWX is not in itself surprising. 24 Indeed, many corporations have found that they can communicate across the nation more economically by writing on and reading a common file accessed via a time-sharing network than by using Telex or TWX. What is slightly surprising to most people is that sophisticated computer mail systems should be cost-competitive.

The 1975 study mentioned above 25 examined how prices will change in the future. The cost analysis focused on the PDP-10 computer, which is used in most current A R P A N E T mall services. On the basis of benchmark tests of a new-generation PDP-10 computer (with a 'KL' central processing unit), the 1977 figures given in Table 2 were estimated. Communication and terminal costs, for the sake of conservatism, were held constant. For 1985, projections were first based on an A.D. Little projection of computer, terminal, and communication costs. 26 Table 2 gives the results of this 1985 projection, under the title 'Method I'.

For the present study, 1985 costs were projected in a second way, which is more intuitive, but more likely to be accurate. The 1985 projection under the second approach is shown under 'Method II' in

2 4 8 T E L E C O M M U N I C A T I O N S P O L I C Y June 1 9 7 7

Table 3. Bell System revenues and expenses, 1975.

Category Billions of dollars

Operating revenue 29.0 Operating expenses 18.8

Current maintenance 5.9 Depreciation 4.1 Traffic 2.1 Marketing and commerical 2.3 R & D 0.2 Accounting 0.8 Employee benefits 2.4 Other 1.0

Taxes 5.1 Interest 2.3 Dividends 2.2 To retained earnings 1-0

Source: Moody's Utilities Manual, 1976.

z7 Future analyses at Stanford will compare computer mail costs with those of interoffice distribution, copying, and kindred expenses. We also hope to test the possibility raised by many researchers that computer mail will affect telephone usage (see David W. Conrath and James H. Bait, 'The computer as an interpersonal communication device: a study of augmentat ion technology and its apparent impact on organizational commiJnication'. Proceedings of the Second International Conference on Computer Communication, Stockholm, August 1 974; and S.J. Lukasic, 'Organizational adjustments to the electronic post office', paper presented at Nat ional Te lecommun ica t ions Conference, San Diego, California, 1974). zs Gilbert J. Konkel and Phyllis J. Peck, Traditional secretarial costs compared to word processing', The Office, Vol 83, No 2, pp 66-67, February 1976.


Table 2. The adoption of a second approach stemmed from a critique of Method I results. According to the Method I projection, the terminal would make up 58% of total costs, and communications another 29%. Only 13% of total costs would be due to the computer and operators. It is my contention that (a) computers will be more labour-saving in the future; (b) marketing and other administrative expenses will grow; (c) intelligent terminals, by editing and storing messages, will slash communication costs; (d) even intelligent terminals will cost much less than the A.D. Little projections; and (e) users will demand more processing power.

One can immediately identify weaknesses in this cost analysis. The most obvious criticism is that reference is to prices, not costs. As shown in Table 3, the US Bell System's annual revenues are over 50% higher than expenses. Moreover, such expenses as billing, marketing, operator assistance, installation, and maintenance, are only partially accounted for in the present analysis. Telex and T W X may be more expensive than current computer mail costs for precisely such reasons. Yet it should also be remembered that commercial computer mail systems actually are being offered at around $1 per message, and that some billing, marketing, and other labour functions are provided for in that price. Computer mail is inherently more economical than Telex and T W X primarily because it is always offered today as one of many software services in general computer systems, and so can share many costs.

Another valid criticism is that only current mail systems were considered, all of which are used primarily to send brief messages within reasonably small user communities. As user populations increase, such services as on-line directory assistance will become necessary. The 1985 computer and communication estimates, in fact, reflect a substantial increase in service power. The present cost estimate should be reasonably good for brief-message services. As noted below, however, there will be a need for computer mail services that can send long documents, handle forms, and do other things that brief-message systems cannot do well. It is not possible to estimate the cost of such systems, but it should be noted that if longer documents are prepared on a word processing system, the cost of distributing the document via computer mail will be quite small.

Another problem with my analysis is that I have been comparing computer mail costs with postage. But, as noted below, an analysis of computer mail traffic indicates that computer mail may be used most often to replace interoffice and even intraoffice mail. 27

All this, however, pales before the fact that relative user labour expenses for various modes of message transmission have not been considered. An analysis by Konkel and Peck 28 indicates that labour charges to research, compose, and type a business letter represent $8-16 out of a total cost of $8.23. Only seven cents of the cost is due to materials and equipment amortisation. Whether computer mail will be cost-effective will depend on how it affects labour costs and, ultimately, on how it affects overall organisational effectiveness.

To summarise, it is very likely that computer mail systems for brief messages will be able to reach prices comfortably under $0.50 per message. Such low prices can certainly be met if mail programs are offered as parts of general computer systems. The costs of mail systems that can handle long messages or complex communication, however, is an open question.



Laek of customer awareness

It is said that one Canadian time-sharing network saw the 1975 postal strike as an opportunity to market its mailbox program. It placed advertisements in several prominent newspapers and magazines, then waited for the orders to come in. It received only one inquiry, from a woman who wanted a package delivered to her son.

While this story may be apocryphal, it underscores what I have constantly observed when talking with line managers: a perception of communication needs, yet no knowledge of computer mail. While rapidly falling computer prices have generated much excitement on the supply side, this same suddenness has made computer mail an invisible revolution for potential users.

It is possible that a flowering of computer mail software, aided by strong publicity, will quickly bring a general recognition of computer mail. But that proposition is suspiciously optimistic. It seems more likely that growth will be very painful at first and remain rather painful even when (and if) rapid growth does set in. We can only guess today whether computer mail will be a universal tool for office workers, like the telephone and the convenience copier, or whether its use will be limited and sporadic, like the loud-speaking telephone and the facsimile machine.

One obstacle that may slow computer mail for years is purchasing. Telecommunications and computers together total only 1 to 2% of the expenses of a business, so their purchase tends to be routine and placed at low levels in the corporation. It is very difficult for a radically new service to be embraced by either the telecommunications buyer or the electronic data-processing (EDP) staff. Moreover, there may be squabbles among telecommunications managers, EDP staff, and the Telex/TWX/facsimile priesthood over control of the new medium.

29 Personal communication with E.C. Johnson, General Manager, Electronic Services Division, United States Postal Service, 2 April 1976. zo Roger Pye, 'The end of the journey to wo rk : fac t or f ic t ion?" Repor t P/741 54/PY, Communications Studies Group, University College, London, 1974. ~ Ford, op cit, Reference 20.

Current systems and organisational needs

A major design issue is whether current computer mail systems meet the real needs of organisations. Perhaps a simple, although ultimately incorrect, way to rephrase the issue is whether current mail systems can at least handle current paper flows well.

From the above discussion of potential market size, one major deficiency of current computer mail systems is obvious: they do not handle forms, at least not easily. Future mail systems must handle forms and correspondence in an integrated manner.

A second point is that virtually all current mail systems can handle only brief messages easily. Long documents, for a variety of reasons, are usually handled separately and inelegantly. But in US Postal Service facsimile tests, the average transmission was 3.5 pages. 29 Similarly, in Pye's analysis of postal and interoffice mail the average item was 3-4 pages. 3° In both cases, one half to three-quarters of all items were a page or less, but longer items brought up the average. Similarly, Ford 3~ found that although 48% of the items copied on convenience copiers were one page long, one-page documents accounted for only 10% of the total volume.

Interestingly, the oldest sophisticated mail system, NLS Journal Mail, which was part of a general office automation system developed under constant user feedback, evolved both a forms system and


3z J.C.R. Licklider and Robert W. Taylor, 'The computer as a communication device', Science and Technology, April 1968. 33 Robert W. Taylor, 'On the relation of interactive computing to computing science', Proceedings of the IEEE, Vol 63, No 6, June 1975. 34 Douglas C. Engelbart, Augmenting Hum. an In te l lec t : A Conceptua l Framework, Stanford Research Institute, Menlo Park, CA, October 1972. 35 F.J. Corbato and V.A. Vyssotsky, 'Introduction and overview of the MULTICS system', Proceedings, Fall Joint Computer Conference, Las Vegas, Nevada, 30 November 1965, quoted in Mathison and Walker, op cit, Reference 6. ze Richard M. Cyert and James G. March, A Behavioral Theory of the Firm, Prentice- Hall, Englewood Cliffs, N J, 1967. zz Thomas J. Allen and Peter G. Gerstberger, Criteria for Selection of an Information Source, Report No 284-67, Alfred P. Sloan School of Management, Massachusetts Institute of Technology, September 1967. 38 FCC regulations extend only to services that are offered for hire.


elegant techniques for handling long messages. The forms system is too new to evaluate, but the Journal system is over six years old and offers interesting data. In one sample studied, while 73% of all items were 170 words or less, the average transmission was just under two pages. It is also interesting to note that 15% of all Journal transmissions were sent to a group of individuals, most often to a whole office, reminiscent of copier usage. Finally, it is interesting to note that at least 6% of all transmissions were entered 'for the record', suggesting that journal mail is used in part as an on-line filing cabinet.

I have emphasised that the computer is more than a switch, yet I have talked only about extending the computer to handle rather routine aspects of message composition and disposition. Licklider and Taylor 32 and Taylor 33 have suggested a much broader paradigm for the computer's role in communication. They view an interaction episode not as the exchange of a series of messages, but as a process in which two people or a group of people attempt to build a common understanding from their idiosyncratic ideas, or, as Licklider and Taylor put it, to build an external model from the models in their individual heads. The computer, they argue, can go beyond the exchange of messages, to assist in the building of models - a task for which it is ideally suited. Similar reasoning went into the conceptual design of NLS Journal Mail 34 and some other systems. 35

But no one has demonstrated a way to get groups to engage in deep modelling, and even computer systems designed with lofty intentions of joint model building have ultimately been used rather mundanely. 'Information apathy' studied by Cyert and March, 36 Allen and Gerstberger) 7 and others suggests that, if computers are to augment the intellectual aspects of human communication, designers will have to abandon the rationalistic paradigms of management information systems (MIS) and kindred tools and squarely face the complex microprocesses of organisational decision making.

Policy issues In the end, public policy may be more decisive than either economics or demand in deciding the fate of computer mail. Although computer mail regulations as such do not exist, there are already several potentially explosive issues and a number of preliminary regulatory skirmishes.

So far, most regulatory activity has been in the USA, arising from the two Computer Inquiries of the Federal Communications Commission (FCC). The First Computer Inquiry, which lasted from 1967 to 1971, resulted in FCC regulations that effectively require all for-hire 3s computer mail operators to be regulated as common carriers, although the regulations do not explicitly mention computer mail.

But the First Computer Inquiry's rules grew obsolete because of the growth in size and complexity of computer technology, and in late 1976 the FCC opened the Second Computer Inquiry. While the rules proposed in the new Inquiry have brought some technical improvements, one problem remains as vexing as ever. FCC rules require computer mail operators to be regulated as common carriers, but, at the same time, limit their ability to offer the sophisticated mail preparation and disposition tools that are precisely the features that make contemporary computer mail services attractive.



a9 President 's Commiss ion on Postal Organizat ions, op cit, Reference 14.

All computer mail services now offered for hire in the USA are in violation of FCC rules, although TYMNET's planned mail service will presumably be in compliance. While this quasi-illicit status has done little to stunt software development, it has driven computer mail underground. Suppliers' reluctance to discuss their services has made computer mail almost invisible, playing havoc with the orderly emergence of a market, although not stopping it completely, and, bringing things full circle, making the FCC unaware of the implications of its regulations for computer mail.

In other parts of the world, regulators have made few rules affecting computer mail, but there have been two major exceptions. The British Post Office is now attempting to work out regulations that will allow US time-sharing networks to be used in the UK but will, at the same time, preserve the lucrative Telex service in the country. Canada, in turn, has a considerable study programme on computer mail, although it has focused primarily on teleconferencing.

The US and UK experiences point to an ominous trend that could permanently distort development: the separation of computer mail's transmission facilities from the total corpus of office automation services, of which computer mail is most naturally a part. This could not only harm the computer mail industry but could sour the development of office automation, which will depend heavily on the presence of communication tools to transfer information back and forth once it is created or processed. Since nearly half all workers in industrialised nations hold office jobs, the economic repercussions of a slowdown in the growth of office automation could be enormous.

In the near future, computer mail is likely to come to the attention of telecommunications agencies in most countries. But not only telecommunications agencies will be involved in regulation. Postal authorities, who are now facing crippling price rises and record deficits in virtually every country, are already concerned that electronic funds transfer systems (EFTS) may draw away as much as 60% of first class mail and 40% of all mail? 9 Computer mail would add to these problems, but it seems most likely that postal reactions would hurt computer mail most if EFTS regulations directly or indirectly included the system under some extension of the postal monopoly.

If computer mail does survive telecommunications and postal regulations intact, many problems will remain for policy makers. Privacy, confidentiality, and security may create regulatory restrictions. And if regulators do allow competition to flourish, some form of mandatory interconnection regulations will certainly be needed.

In the present discussion, only intranational computer mail problems are considered. But international record traffic already provides roughly 20% of the revenues of international telecommunications carriers, and international computer mail traffic is likely to be considerable, provided international technical and tariff agreements are successfully negotiated.

Conclusions

Computer mail has a great deal going for it: a favourable economic outlook, a huge potential market, and weakening postal opposition. To tap this market, considerable design evolution will be required, but


~°The CB radio system is like an open party line, into which anyone can enter. It has been growing explosively, and about o n e US household in ten has a CB radio. 41 In addition, when Bell Canada recently tested its computer teleconferencing system, CMI III, the most active session was the 'Graffiti ' conference, in which users merely exchanged anecdotes (personal communication with Gordon Thompson of Bell Northern Research Laboratories). 42 j . Vallee, R. Johansen, and K. Spangler, q'he computer conference: an altered state of communication?', The Futurist, June 1975, pp 116-121. 43Alfred D. Chandler, Jr, Strategy and Structure: Chapters in the History of the Amer ican Indus t r ia l Enterprise, Massachusetts Institute of Technology Press, Cambridge, Mass, 1 962.

Ibid. 4s Cyert and March, op cit, Reference 36.


software is already evolving rapidly, and design changes should pose no intractable problems. But there is also the strong possibility that regulatory agencies will artificially separate computer mail from the general office automation systems, with which they must ultimately be integrated. Although growth seems certain to come, it will not come immediately, nor will it come easily.

In this article, I have looked only at the uses of computers in organisational communication, to the neglect of computer-based systems for private citizens. This focus stems from the assumption, shared by most analysts, that organisations will have the resources to adopt computer mail before private individuals. Yet it is conceivable that the rapid growth of personal hobby computers in many countries will give rise to something akin to the recent growth of Citizens' Band (CB) radio in the USA. 4° Perhaps it will be in the form of 'CB computer mail' that computers make their first major inroads into communication. While this idea may seem far-fetched, it should be remembered that many computer mail uses on the ARPANET today fall into the 'hobby' category, albeit on a grand scale. 4~

As mentioned above, several writers have suggested that computer mail is only an early precursor of the computer's ultimate role in organisational communication - j u s t as early number-crunching computers failed to hint at even today's computer uses. But the grandest visionaries have so far suffered complete failures of imagination in their attempts to articulate these hazy potentials in concrete terms or even evocative images. Teleconferencing and a handful of other experimental tools provide some intimations of 'altered states of communication 43 but even these are beggarly dreams, and their implementations have produced only limited insights.

In the end, exotic applications of computers may prove unwanted and even a little silly. But because of the central importance of communication in human collaboration - the typical manager spends roughly 80% of the day in communication 42 - futuristic thinking about computers and human communication may be crucial to future productivity growth. If we design our computer-based human communication systems too ploddingly, we may unwittingly deter or even deny future applications of enormous importance. At the very least, we need to spend more time considering the future.

It must be borne in mind, whenever thinking about human collaboration, that the modern complex organisation with its multidivisional structure is a very new creatiori. General Motors, Sears, Du Pont, and a few others pioneered the modern form only in the 1920s, and it was not until after the second world war that the complex organisationai form began to dominate industry. 44 Because of its youth, the modern organisation is still evolving rapidly, and it seems likely that collaborative (and communication) patterns are evolving too. It is likely, then, that we must understand how complex organisations composed of individuals who work only 40 hours a week and who, by all indications, 45 find it almost impossible to achieve complex interactions in group efforts, are nevertheless successful in adapting to today's highly complex business environment in rather appropriate manners. Only when we understand how complex organisations successfully achieve collaboration will we be able to design computer systems for effective human collaboration.


Documents

The outlook for computer mail