© 2007 Chris Kern
It isn’t obvious today, when anyone can pick up a copy of Microsoft Word or OpenOffice Writer and prepare text in any of the major languages of the world, but it wasn’t so many years ago that writing anything on a computer except in English and a few Western European languages typically wasn’t feasible. So in the early 1980s, when the Voice of America canceled the planned purchase of a mainframe computer system for its English-only central news department and solicited proposals for a distributed computer network to support all 40 of its broadcast language services, the prospects for the success of the procurement were far from certain.
I was one of two “computer guerrillas” who persuaded management to scuttle the centralized English-only system proposed for VOA by the information technology staff of our parent organization, the former United States Information Agency. The other was my colleague Don Barth, a radio technician who had single-handedly installed a primitive network of early microcomputers in our central newsroom. Don recruited me because I had done some computer programming in college during the middle 1960s and had assembled a couple of microcomputer kits as a hobby while working for VOA as a journalist. I’m amazed in retrospect that two amateurs were able to derail the USIA project. But those were simpler times. The traditional bureaucratic rivalry between VOA and its parent agency undoubtedly helped.
At the climax of the Cold War, the majority of VOA broadcasters were preparing radio scripts for the “accurate, objective and comprehensive” news mandated by their legislative charter on a motley collection of electrical and mechanical typewriters, a few of which had actually been manufactured before World War II. Some were almost impossible to maintain: it wasn’t easy to impossible to maintain: it wasn’t easy to find spare parts for a Bulgarian typewriter in downtown Washington. Even when the machinery worked properly, typing many of our languages required both dexterity and the patience to backspace and sometimes raise or lower the typewriter platen so base characters could be combined with diacritic marks to create the composite glyphs that would form the words. In some VOA language services, typewriters were either unavailable or too difficult for the broadcasters to use; radio scripts were written by hand and edited by hand, and the scrawls were painfully—and often quite audibly—deciphered in real-time by the on-air talent.
We were therefore delighted when one of the four bids submitted in response to our request for proposals was a commercial product that already offered word processing in the majority of the VOA languages, at least in prototype, and clearly could be coaxed into supporting all of them. The product was the Xerox Star, a computer workstation with what was at the time a radical design featuring a bit-mapped display, a graphical office motif with folders depicting filesystem directories and icons representing other objects, and copy and move and save and print operators which functioned by selecting one operand with a pointing gizmo called a mouse and dragging it to the location of the second operand on the computer monitor.
Not only that, but the Star workstation was the front end to a remarkably sophisticated distributed computing environment that had most of the major attributes of the modern Internet Protocol. The Xerox Network Systems protocol and suite of services offered central or departmental filing, remote printing, and suite of services offered central or departmental filing, remote printing, electronic mail, authentication, and directory services for all the computers and users on the network. The “IBM PCs” of the era offered vastly inferior facilities on the desktop, and either operated standalone or on networks which were comparatively primitive. Xerox, like all the other bidders, proposed to install PCs in our English central news department in order to maintain a competitive price. But at the insistence of the newsroom representatives on the evaluation team, we selected a contract option to provide Star workstations for the central news staff. (We ultimately wound up upgrading to a somewhat faster successor to the Star product which ran the same software as the Star.) Don Barth coined the name System for News and Programming for the new network; the acronym SNAP had a nice ring to it and, as lagniappe, a “snap” was the term the Reuters news agency used for what we in the United States referred to as a news “bulletin.”
The installation began in 1986. Aside from the United Nations headquarters in New York, there is probably no place in America where so many people from so many cultures, speaking so many languages, are packed into a single building. We figured some parts of the organization would have more difficulty assimilating the new computer technology than others and, with the help of the Xerox federal marketing staff, we prepared a fairly elaborate series of briefings and training sessions to ease the transition. But we couldn’t anticipate all the reactions, and on more than one occasion we were taken completely off-guard.
The first surprise came from the central newsroom, which had contributed two of the six representatives to the bid evaluation team. Nevertheless, the news director was outraged at our choice, insisting that we should have arranged for the development of custom software to meet the News Division’s particular requirements rather than purchasing a general-purpose system which could support all the language services as well as the newsroom. Many members of the newsroom staff chimed in after we made the mistake of deploying a new and somewhat flaky release of the Xerox workstation environment in order to take advantage of updated, higher-performance workstation hardware. An independent newsroom protest was started by a copy clerk who was also a computer aficionado. He circulated a jeremiad—written on the Xerox workstation, of course, and distributed widely as electronic mail through the XNS computer network—claiming no one would ever be able to comfortably use the workstation because of its reliance on this idiotic “mouse” device.
At one point, in desperation, we actually offered—perhaps threatened is more accurate—to reinstall the previous newsroom system, which was based on a generation of microcomputers that antedated the IBM PC and had a penchant for multiple-node crashes. Not surprisingly, the offer was vehemently rejected and we were told to “go fix the problems with the Xerox system.” Also not surprisingly, that was the response we had hoped for. A subsequent and better debugged release of workstation software took the edge off the complaints even though it didn’t entirely stop the grumbling.
Meanwhile, we were proceeding to finalize the word processing capabilities in various languages. A term in the contract gave VOA an opportunity to review the prototype for any language software package that wasn’t already a commercial product. Actually, the Xerox software designers were delighted to gain access as free consultants to professional journalists who were native speakers of the various languages. In most cases, these consultations went smoothly. The Xerox representatives were happy with the suggestions they received on how to make the software more intuitive to native speakers and the VOA broadcasters were pleased to have been able to influence the delivered product.
Not all these meetings were congenial, however. I watched one morning in a makeshift lab as a tall, imposing Hungarian broadcaster sat motionless and listened intently while a rumpled computer scientist from Silicon Valley explained how his software could almost magically combine letters and diacritical marks of the language into composite glyphs, how it was consequently possible to dispense with the “stop keys” and other clunky mechanical contrivances required by a typewriter and therefore write intuitively, concentrating on the meaning of the words rather than the process of putting them on paper—or, in this case, on a bit-mapped computer monitor.
I was observing, but I confess I wasn’t paying much attention. The computer scientist was Joe Becker, who had invented the Xerox Star’s multilingual technology and now was serving as the lead technical designer for its conversion into a commercial product; there wasn’t much I could contribute to the dialog between Joe and his Hungarian interlocutor. Therefore, I didn’t pick up the warning signals. Suddenly the broadcaster lurched up and, pointing with a shaking finger at the “virtual keyboard” on the computer monitor, announced in an angry but still sonorous radio voice: “This machine is an affront to my national heritage.” Hungarian, he informed us, must be typed on a computer keyboard exactly as he typed it on his IBM Selectric typewriter. No other input method was acceptable. And then he stalked out of the room. The consultation apparently had ended.
I was appalled, and a bit embarrassed since Joe’s input method was elegant and it was evident that, even in prototype, it had been finely implemented and neatly integrated into the general computing environment. But the broadcaster was the end-user and I felt obligated to ask the question, no matter how foolish I knew it to be. Would be possible, I inquired, to modify the software so its typing logic mimicked the IBM typewriter? Joe, though obviously taken aback, didn’t hesitate. “It’s an extremely flexible system,” he said, “we can make it as bad as you want.”
A Revolutionary System
We experienced a different kind of adverse reaction to the input method in the Chinese Service. Chinese is an unusually difficult language to type. A reasonable vocabulary for news broadcasting requires the use of 4000-6000discrete Chinese characters, or ideograms. All of these represent individual words and they may also be combined to form compound words. To type Chinese in the Xerox document editor, the user entered the words phonetically with the Roman alphabet, using a method called Pinyin to represent the sounds of the standard dialect of Chinese, Mandarin. Of course, that only worked if the typist spoke standard Chinese rather than one of the many distinctive regional dialects. But that was no problem in the VOA Chinese Branch, where speaking Mandarin was a job requirement.
However, Pinyin had been adopted on the Chinese mainland as part of an effort to standardize pronunciation, and some of our Chinese broadcasters who had emigrated from Taiwan denounced the Xerox typing method for its “communist” origin. Their indignation was compounded by a decision to print all radio scripts in the simplified Chinese characters used on the mainland rather than using the more ornate traditional characters used on Taiwan. The new Chinese Service chief, a young foreign service officer named Tony Sariti, had just returned from a diplomatic tour of duty in Beijing. Tony, a fluent Chinese speaker, was concerned that VOA’s Chinese programs sounded odd to a mainland audience because of the outdated vocabulary of our émigré broadcasters. He was determined to recruit broadcasters from the mainland to modernize they way VOA sounded in Chinese—and since it was difficult for people from the mainland to adapt to the traditional characters but easy for those from Taiwan to read the simplified ones, he opted for the latter.
The reaction probably wouldn’t have been much more intense if we had issued every Chinese broadcaster a copy of Mao’s Little Red Book. Some issued every Chinese broadcaster a copy of Mao s Little Red Book. Some members of the staff, mostly middle-aged and older men, boycotted the computer altogether, insisting that they would continue to write their scripts by hand. Their resistance began to dissipate, however, when Tony hired an extremely attractive young female contractor who was an expert Chinese typist, and announced that she was available to tutor any member of the Chinese Service who was interested. I knew we had turned the corner when Tony came running into my office one afternoon and announced with a big grin that “the guys who were the worst holdouts are lined up three-deep waiting for their computer typing lessons.” And, of course, producing Chinese scripts on the computer had distinct advantages over writing them by hand that no one in the service could deny. Among other things, with clean copy to read in the studio, the broadcasters could speak at a normal rate on the air instead of the artificially slow pace they had been forced to adopt while trying to decipher the handwriting of the writers and editors.
In some other language services, there was no objection to the software, per se, but adoption of this new method of preparing radio scripts was slow. Arabic, for example, seemed like a perfect candidate for word processing. The Xerox software could write from right to left, and was able to automagically perform the required transformations in the shapes of Arabic letters depending on their initial, medial or final positions within a word. But many of the Arabic broadcasters were accustomed to writing by hand, then turning over their copy to a professional typist to produce a finished script. Learning to type was difficult, especially for the older members of the staff. Eventually, though, most were able to do it.
What actually turned out to be a more serious problem for the Arabic broadcasters was the appearance of the text. It was readable enough for use in a radio studio, they told me, but they wanted to create an audience newsletter and a program guide, and couldn’t use the existing font in printed materials. I phoned Joe Becker of Xerox at his office in Northern California and asked his reaction to the Arabic Service’s complaints. To my surprise, he emphatically agreed with the broadcasters. “That font is really ugly,” he explained. “You can think of it as the equivalent of dot matrix printing. Don’t inflict it on your audience. The product manager is trying to license a professional Arabic typeface. Wait for that.”
We did. It took about a year to materialize, if memory serves, and while it never looked as nice as the type in the books and magazines the Arabic broadcasters showed me as examples, even to my uneducated eye, it was presentable enough for our purposes. Meanwhile, I had learned a useful lesson about the difference between my native language and those with a more deeply-rooted calligraphic tradition. Even at its best, modern English text can best be described as utilitarian. It is optimized for readability, not beauty. But the appearance of the written word was as important as readability for the graceful, cursive languages of the Middle East and Asia. Since most of the VOA broadcast services wanted to distribute written text to their listeners in the form of program schedules or audience newsletters, we needed to consider paying extra for presentation-quality fonts.
Although Xerox had most of VOA’s major languages bundled into the Star product or available in prototype by the time we signed the contract, technical issues developed as we moved into some of the more “exotic” languages. Internally, each letter or other component of a language needed to be represented by a code. Although our broadcasters would never see the codes, the internal representation of each language needed to match the existing character-encoding standard in the area of the world where the language was spoken if it was going to be possible to exchange text electronically. But different countries, and even different regions within a country, sometimes had strongly-held opinions about the proper way to code the same language. These had to be evaluated before a choice could be made about which was the best multinational fit.*
A few languages required functionality that did not exist in the software we began deploying in 1986. The ligatures that connect the letters of some South Asian and East Asian languages, for example, had features that differed from the ligatures already supported by the software in the commercial product. In addition, diacritic markings in several East Asian languages were more complex than those of European languages. Surprisingly, even Vietnamese—written in the Latin alphabet—required a significant redesign of the rendering software to accommodate its elaborate diacritics. Typing languages with a “virtual keyboard” on the computer monitor also proved too difficult for many broadcasters and slowed up the work flow for most of the others. We had to contract for physical keycaps to be designed, printed, and inserted on the standard Xerox keyboards.
Thought we hadn’t been able to anticipate all the details, we had known from the beginning of the project that some languages would be difficult to implement, and had developed a priority list for language deployments based on two criteria: the maturity of the Xerox language package and the importance of the language to the Voice of America. The central newsroom and worldwide English service went first. The big VOA Russian and Ukrainian Services, both working in the Cyrillic alphabet, were the obvious choices for the initial foreign language installations. By the time we tackled Russian and Ukrainian, Xerox had wrung out the initial bugs in the operating system software for the new workstation hardware which had caused the debacle in the newsroom. Cyrillic script was no more complicated than English to input or render. All the Russian and Ukrainian broadcasters were accustomed to typing their scripts. We expected the deployment to go without a hitch.
And so it did—at first. The Russian broadcasters in particular had no difficulty adapting to the computer, and quickly exploited the Xerox network file service to develop a sophisticated copy flow regime. Managers and staff alike immediately took to electronic mail—still a novelty in the middle 1980s—and the on-air talent appreciated being able to carry “clean copy,” unmarked by handwritten edits, into the broadcast studio. After I was selected to manage the VOA Computer Services Division in 1986, I had lured a veteran editor named Al Riddick away from the central newsroom to serve as the head of my customer support operation, figuring a journalist would have both credibility with the broadcasters and the ability to help them figure out how best to adapt the technology to a news environment. Al had developed an interest in the integration of computing technology into the editorial process, and he worked closely with the Russian broadcasters to get their operation running as smoothly as possible before moving on to the next set of language services. Then all hell broke loose.
One morning Al Riddick walked into my office with a grave expression on his face and informed me some members of the Russian Service staff were extremely concerned about the “health effects” of using computers. Al didn’t know the details, but said he had accepted an invitation for the two of us to discuss the complaints in an open meeting that afternoon. We talked it over, and concluded the problem must have been the institutional florescent ceiling lights used throughout the building. These were unpleasant enough under any circumstances: harsh and glaring. When they strobed against the refresh rate of the Xerox computer monitors, which might happen depending on the angle of the monitors, the effect could indeed be unpleasant. A call to our facilities people elicited a promise to inquire about the feasibility of upgrading the office lighting, so at least I figured I was prepared with a reasonable response when Al and I faced the Russian broadcasters later in the day.
Sure enough, the first question was about the pathological interaction between the flickering florescent lights and the computer monitors. I delivered my rehearsed answer and it seemed to satisfy the questioner. There was silence for a moment and I wondered if the meeting was over. I was feeling pretty smug. But then another broadcaster introduced herself as a physician (it later was explained to me that “nurse practitioner” would have been the American equivalent), and began reading a description from a medical textbook of the awful biological consequences of exposure to high doses of ionizing radiation. The symptoms included, she said with rising agitation, blistering sores, hair loss, internal organ damage, and ultimately even death. All that was true, I conceded, but computers produced electromagnetic radiation—and not very much of that—not the kind of high energy particles she was reading about in the medical textbook.
They do produce ionizing radiation, she insisted.
They don’t, I said.
The Patient Presents
Our dialectical impasse was broken by a short, stocky older man named Iosif Perl (not his real name) who pushed his way up to the front of the crowd and announced in a loud voice that he had been rendered sexually impotent by the poisonous rays from the computers, and expected the United States Government to pay restitution. Not a little taken aback, I lamely made another speech about the difference between the high energy radiation the “physician” was describing to us from her medical textbook and the type of radiation that might be emitted at very low levels from electronic equipment. Mr. Perl was not impressed. He not only demanded money damages, he said, but these must include punitive damages because if we didn’t know in advance that serious illness would result from the introduction of computers, and therefore were guilty of malice, then we obviously had been criminally negligent.
As he spoke, his face grew redder and redder, his voice became louder and louder, and he began gasping for breath. I decided we had lost this round and, muttering something about the need to confer with the agency’s safety officer, headed for the exit with Al following close behind me. Our escape was blocked by a young woman. “I want to apologize for my colleagues,” she said. “Some of them tend to become hysterical.” I thanked her for saying that, and said I just hoped Mr. Perl wouldn’t have a heart attack. “Oh, no,” she sighed, “Mr. Perl acts like that all the time—and unfortunately he never has a heart attack.”
While I tried to figure out how to persuade the Russian broadcasters they had nothing to fear from radiation, Al confronted the problem more deviously. And effectively. He somehow managed to plant a rumor in the Russian Service that the Ukrainian broadcasters down the hall were sneering at the ignorant Russians for their ridiculous concerns about radiation. A few days later, he gleefully reported that a Russian broadcaster had whispered to him that the loutish Ukrainians were too stupid to employ the computer network with the sophistication that had been demonstrated in the Russian Branch. There was no more talk of radiation.
Shortly after that incident we hired Walter Torrance, a young former Air Force specialist with an affinity for languages, as our full-time liaison with the broadcast services. Though Walt had no prior journalistic experience and was not at the time a computer techie, he quickly developed an understanding for what the broadcasters needed to do their work and mastered the nuances of word processing in many languages. To my astonishment, before too long he had managed to teach himself to speak at least a few words of every VOA language. It was largely through his efforts that our difficulties in some of the initial language services were not repeated as we introduced multilingual word processing to the rest of the organization. (Walt worked with the VOA broadcasters to produce the translations of the Voice of America Charter which accompany this article.)
Computer scientist Douglas Hofstadter’s Law states that “it always takes longer than you expect, even when you take into account Hofstadter's Law.” The multilingual computer system we envisioned in the early 1980s would take us to the end of the decade to fully implement. By the time a popular television program called The Computer Chronicles came to VOA in 1991 to report on what we had accomplished, we were already planning a successors system based on commodity personal computers running on an Internet Protocol network. Xerox, which had designed and manufactured the remarkable computing machinery and software that made our project possible, was meanwhile being accused of “fumbling the future” for failing to effectively market its inventions. Maybe that’s true, although I think mostly hey were just ahead of their time. Fortunately, they came at just the right moment for the Voice of America.
* This problem was soon to be solved with the invention of the Unicode standard by Joe Becker of Xerox and Lee Collins of Apple, although securing international consensus would take several years.