7.3  "The Origins of Email - Encyclopedia Article"

By Alan F. Kay

© 2000, all rights reserved
June 10, 2000 

 Abstract

It is commonly believed that email originated on the Internet.  Not so.  Some email systems were used by small segments of the population before Arpanet, the precursor to Internet, sent its first messages from one computer to another in late 1969.  Also there were forms of email commercially available many decades before the Internet was open to the public.  Early email had some good features and some of these unfortunately seem to have been lost on the Internet as it evolved.   This early history, with surprising twists and turns, is almost unknown. Somewhat similar to the history of email, the pre-Internet development of e-commerce, is described in a peer-reviewed Global Focus article[i] (see 6.2 The Birth of E-Commerce).  The near parallel developments of e-commerce and email are two good examples of how the expenditures of huge amounts of money to capture burgeoning markets play a key role in distorting public understanding of how important innovations arise and flourish.  This paper traces the early history of email with that viewpoint in mind.

---

It’s common knowledge that Arpanet, which later became the Internet, was designed to allow widely dispersed computers owned or subsidized by the Department of Defense (DOD) to share programs and data.  This enabled DOD to manage the taxpayers’ money responsibly by more cost-effectively handling its computing needs and resources, both human and otherwise[ii].  DOD computers of different eras, sizes and manufacturers were often one-of-a-kind and viciously incompatible.  The goal of the developers of Arpanet was to permit any properly authorized user, sitting at a network node (a DOD computer), to access any other node in order to readily upload and download programs and data, even if the user was otherwise unfamiliar with the computer at either end.  Simple message transmission/reception, which came to be known as email, was an incidental, minor but useful, byproduct for government business.  Personal business was avoided as improper.  Funded for many years by government and universities for research, education and other large purposes, the Internet never had an extra charge for email messages.  

That “incidental” email capability became the key factor of the Internet’s fabulous success.  Compared to postal mail, telephone, fax, or any other way to get a message from point A to point B, nothing was as cheap as the essentially zero-cost email.  In time, many people on limited budgets became Internet users, for its various advantages.  When money is tight, as it almost always is, few held out against free mail for long.[iii]   One of the attractions was, of course, the WorldWideWeb, but the principal outcome of the presence of the Web was the conversion of users of email into customers for e-commerce.  The presence of millions and tens of millions of Internet users produced a growth spurt for business-to-consumer (B2C) e-commerce.[iv]  Until the 90’s, Federal government investments of tax dollars in the Internet were based on performing scientific, academic, and research functions, not mail or commerce.  It was only when the Internet was opened to all that email and e-commerce flourished. 

First a little history.  Many early e-commerce companies, were actually more message transmission oriented than the Internet itself, and therefore had some form of email.  Until the computer came of age, the innovations with the greatest impact on communication – the telephone and the television were unlinked.  The first commercially available service provider that linked its central computer via the telephone network to its remote customers with full screen TV and keyboard terminals was a start-up company, AutEx, which began revenue-generating operations on August 1, 1969, and was soon profitable.  Believe it or not, prior to AutEx, customers with TVs hooked to a remote computer belonging to a supplier or a third party did not exist.  The first AutEx network and each that followed in the next few years linked companies in some industry to their suppliers and/or customers with transaction-oriented software that made AutEx’s central computers “marketplaces.”  AutEx called them marketplace systems.  Today it’s called “e-commerce.” 

The AutEx “supermessage”, with key email features, was in use and generating revenue by August 1, `69, some months before the predecessor of the Internet sent its first test messages from one computer to another.  The supermessage was available and heavily used in conjunction with AutEx’s e-commerce industry-wide systems.  For reasons explained below, it was designed to look and feel like one of the many other message types that AutEx offered for industry-specific e-commerce.  These e-commerce message types were in tightly coded, fixed format.  To simplify the description of the supermessage capability, it is worth first explaining a typical e-commerce message.  The interest message, the most common message of AutEx’s block trading system, for example, is illustrated by B L IBM, L1 (meaning, with keystrokes shown in bold) “willing to Buy a Large Amount (over 5000 shares) of  IBM stock, message sent to List 1 (all institutions).”  Interest messages, sent by broker/dealers to financial institutions, would flash on the recipients screens in a ticker-like display, neatly formatted in columns showing the latest messages with Buys on the left and Sells on the right.  In the stock field, illustrated by the IBM example, in addition to stock symbols, bonds with coupon and maturity date were accommodated.  Each message could be sent to any subgroup of users by use of system-wide lists, like L1, plus the broker/dealers own lists similarly coded, e.g. L10.  Individual users could be added or deleted by + and/or – signs; as illustrated by the  distribution instruction “L10 + 363 + 221 – 119” meaning “send only to users on our list 10 with user 119 deleted and to users 363 and 211.  The supermessage allowed a similar list-and-user-number based distribution instruction and was displayed with an up-to 10 character title and an 8 character user acronym serving as a header to free-form text of 8 lines, each up to 40 characters in length, created by the user assisted by an editing capability.  If the address was a single user, the message was re-categorized as a “direct message.”

The supermessage appeared within two seconds on all screens specified until replaced by a subsequent message.  In addition it was retrievable by recipients three different ways:  (1) by time/date period, (2) by key words in the title (use of wild cards allowed), and (3) by sender.  Of course, the supermessage was limited by the small numbers of users on a single system (typically under 1000) and by the 320 character message length.  Beyond delays of a few seconds caused by transmission and processing protocols both by AutEx and by Internet email, AutEx  supermessages experienced no further delays that are found routinely on Internet email, corresponding to the elapsed time between message sent and recipient’s request to receive incoming mail.  This delay is sometimes zero but is typically hours and often over a day.  I think a fitting description of any mail system, postal or electronic, with that kind of delay is “snail mail”.  

Retrieving a supermessage previously sent did not require filing it first.  Today’s email systems are primarily quite elaborate systems for filing, storing, searching, archiving, moving, and deleting mail.  There is no limit as to how sophisticated a filing system one might want to create, but such systems are still limited to easy accommodation only of email.  Without considerable reorganization of other systems for handling received/sent information, an email filing system cannot accommodate both sent and received, and both copies and/or originals of (a) voice mail, (b) voice or recorded phone messages, (c) postal letters, (d) fedexes and similar high-end mail, (e) internal or private paper-messages (pink slips, etc.), (f) other email received/sent through independent accounts, and (g) generally speaking, faxes.

To flesh out the history of email, it is helpful first to retrace a little more of the early history of commercial computing that I both observed and participated in and which led to the start of AutEx. 

As a PhD mathematician and engineer working in the defense industry beginning in ’51, I had limited access to the few “giant” experimental computers then under development in universities and government facilities and, if commercially offered, like the ENIAC, were priced at over a million dollars.   By ’54-’55, spec sheets appeared for general-purpose digital computers with prices in the $100,000 per copy range.  Four companies offered these new machines with the functionality of the personal computers of the early ’80s, but each would occupy a full-size, air-conditioned room, was much slower, labor-intensive (vacuum tube-based), and had severely limited capacity.  The four companies were all out of business within a few years.  A NYC based-firm, TRG, where I was working in ’55, bought one such machine, the Alwac computer, and used it effectively for many years for scientific and engineering calculations.  IBM, which was to dominate the business for decades, was not one of the four companies.  Trying to protect its punch card handling equipment, IBM was quite slow to become a provider of new types of electronic digital computers.

This was all a decade before Gordon Moore, co-founder of Intel, in 1965 when still at Fairchild Semiconductor Co, wrote a paper predicting a fabulous long-term drop in memory cost, spawning what has come to be called Moore’s Law[v], often phrased as a doubling of performance and/or halving of costs of computers every 18 months.  In the dozen years before Moore’s paper, there had already been about six or so doublings of cost/performance, so the concept was recognized by those following such things even during ’51-’65, the early pre-historic era.  (Since almost everybody thinks that both email and e-commerce started on the Internet in roughly the last decade, I consider all developments of these capabilities before 1989 as pre-historic.) The publicly available history of computers, as well as the millions of computer professionals now working, seem unaware that even in the early prehistoric age the doubling was apparent.  It did not begin in `65.

            TRG was sold to Control Data in 1963 and it was time for me to take on a new challenge.  My experience in using big computers on engineering problems made it logical for me to believe that computers were going to be very big in business.  In the mid ’60s, the only business applications of computers were based on the huge mainframes of large companies, which were slowly putting routine clerical functions like general ledger and inventory into large batch processing operations.  But that did not interest me.  I was excited by the potential of “real-time, widely distributed-network systems”, then novel, today ubiquitous.

The AT&T Hammerlock

Using conditioned lines leased from the telephone monopoly, AutEx’s service was based on circuits each linking some ten or twenty customers continuously on-line to its computer.   The computer initially polled users at 1800bps (and later at higher data rates) and at each poll accepted incoming messages and delivered outgoing messages.  The standard was not more than a two-second response delay, like a fast postman walking all the routes in town, receiving and delivering messages at each house and covering the whole town every two seconds.  To accomplish this AutEx needed to link its central computers to the existing telephone network by the end of 1968.  The AT&T monopoly would not tolerate competition in the point-to-point communications business.  The problem was serious.  To reduce the chances of AT&T refusing to service AutEx, we minimized message transmission capabilities and hid them behind e-commerce B2B services, as illustrated above by the supermessage seeming to be a simple extension of the interest message.  In fact, it turned out that AutEx had few problems with AT&T and these were beautifully handled in the critical early days by management persons familiar with the telephone network and the culture of AT&T and its subsidiaries. 

Until the early `70s every voice telephone connection from point A to point B and every “private” line or line leased to a non-Bell user had the full bandwidth of a voice line.  If such a line carried only occasional short messages, the cost of the line when allocated on a per-message basis became extremely large. AutEx’s computers, by more fully utilizing the carrying capacity of the still uncomputerized telephone network, had also made a breakthrough in lowering intrinsic message delivery costs.  What this implied is that AutEx could in principal offer communication per-message services commercially for far less than the phone company itself was then charging.  In some cases, AutEx estimated a cost per-message improvement of over ten to one.  Looking ahead to a much later time when the Internet accommodated commercial messaging services, the packet-switching concept was used with a cost-per-message effect similar to the polling method used by AutEx.  Little AutEx’s capability had no negative impact on common carriers, controlled as it was by its dependence on AT&T.  The effect of the Internet, on the other hand, as it grew from a million to tens of millions of users, did begin to force common carriers to lower their per-message rates.  As the Internet grew the pricing plans of all communications carriers had to repeatedly cut per-message charges to compete with the Internet standard of virtually free messages. 

AutEx assigned users to one of the many circuits it leased from the telephone company, each frequently re-configured by AutEx to cover its expanding industry and customer lists with adequate service at minimal cost.  The method of polling circuits loaded by slightly smart terminals[vi] installed and maintained by AutEx at customers’ locations is a communications technique not at all original with AutEx.  It had been used earlier by some large companies, like IBM and AT&T itself, but at the time only for experimental in-house networks.  When AutEx offered supermessage capability as a part of its marketplace systems, no one else was offering a general purpose, electronic messaging service with intrinsic cost per-message greatly reduced by circuit polling technology.  Today only Local Area Networks (LANs) generally use this method.  For linking business to any segment of the general public, the zero per-message cost now makes anything but using the Internet too expensive.  Since the cost-effectiveness or the efficiency of circuit polling compared to packet switching is approximately the same, it is not the technique of packet-switching per se that justifies a business model without extra revenue for message volume.  No per-message charge, sometimes justified by companies on the Internet as due to the “savings” of packet switching over previous technologies is misleading.  The concept is essentially a marketing gimmick. 

One effect of all this was that AutEx turned away from the possibility of becoming an alternative long distance carrier, as such entities began to appear on the scene and prosper despite their questionable legality for a time that ended even before Judge Harold Greene broke up the Bell monopoly.  Even though I knew that it could have been an enormously profitable business, I was not prepared to fight AT&T, as say MCI chose to do.  High legal hurdles kept AutEx out of the telecommunications business, while the much later Internet e-commerce missed this era entirely.  Its battles had been fought and won by others. 

Pre-historic vs. Modern, Odd Similarities in a Field of Differences

Since Arpanet was clearly a research project of a government agency, it did not run into the difficulty with the monopoly owner of the telephone network, the still childless MaBell, which constrained the commercial development of AutEx and pushed it down a different evolutionary path.  By the time the Internet had any commercial communications attributes, MaBell first had been forced to open its networks to computer-controlled communication and later to divest the baby Bells.

The story of the development of Arpanet, explained in delightfully clear, non-expert language by Hafner and Lyon[vii], was familiar to me from the beginning.  There was little competitiveness or secrecy between AutEx and Bolt, Beranek, and Newman (BB&N), the initial developers of Arpanet.  The Chief Executive Officer (CEO) of BB&N, Sam Labate, joined the AutEx Board of Directors after AutEx went public in `72.  When Sam retired, his successor CEO, Steve Levy, was Sam’s replacement on the AutEx Board.

There was one remarkable similarity between prehistoric and Internet e-commerce.  Accompanied by the bitter retreat of the “free Internet” netizens and hobbyists movement, many of whom had once hoped that the Internet would become the unfolding nervous system of the evolving human family, B2C flourished.  Many Internet e-commerce businesses found user advertising, the ubiquitous banner ads, to be the first source of revenue, sometimes still their primary source.  In the case of AutEx, by 1970 ingenious users on their own had learned to lay out supermessage text, which at the time provided just 8 lines of 40 characters each, with low-resolution cartoon figures, logos, etc., producing the “joke of the day” and similar advertisements with the often serious and successful marketing purpose of raising the visibility of their wares.  Both prehistoric and Internet e-commerce over twenty years apart – almost from their first days on-line – found users looking for advertising as a means to distinguish themselves from their competitors. 

The capabilities of prehistoric email went far beyond the AutEx supermessage.   Email was embedded in the services of other providers, such as Compuserve, Prodigy, the Well, and the Source during the ‘70s and ’80s.  The 320 character limitation disappeared, and along with longer messages came the need for more comprehensive editing, i.e., word processing.  Bulletin board systems were offered by dozens of low cost service providers often operated by one or two people working out of a room at home. The improved service sometimes focused on more sophisticated distribution, still limited to text messages.  “Computer conferencing” serviced hierarchies of multiple chat rooms for geographically diverse groups that needed to communicate in different and complex ways within its various subgroups, using facilitators and trainers, and requiring some limitations on who participated in and/or led which subgroups.  For a few years in the mid eighties I helped Participation Systems, Inc., develop and promote its computer conferencing ideas.  By the late ‘80s most email services linked up with the Internet – they just could not beat that zero per-message price – and so they had to live with Internet limitations, design and operational problems.  Thereafter, to my knowledge, the history of email seems to have been merged into the history of the Internet.