Where Wizards Stay Up Late: The Origins of the Internet

by Katie Hafner

  On May 12, 1977, Ken Pogran, John Vittal, Dave Crocker, and Austin Henderson launched a computer mail putsch. They announced “at last” the completion of a new mail standard, RFC 724, “A Proposed Official Standard for the Format of ARPA Network Messages.” The standard they were proposing contained more than twenty pages of specifications—syntactical, semantic, and lexical formalities. The RFC explained that the receiver of a message could exercise an extraordinary amount of control over the message’s appearance, depending on the capabilities of one’s mail-reading system.

  In the days after the publication of RFC 724, the computing community’s response was at best cool to the new protocol. Alex McKenzie of BBN was particularly outspoken. Postel, who had been a defender of the old RFC 680, was the least impressed by the new proposal. He came down hard on the assertion that this was to be an official ARPA standard. “To my knowledge noARPANETprotocol at any level has been stamped as official by ARPA,” he said. “Who are the officials anyway? Why should this collection of computer research organizations take orders from anybody?” There was too much emphasis on officialism and not enough on cooperation and perfection of the system. “I prefer to view the situation as a kind of step-by-step evolution,” he said, “where documents such as RFCs 561, 680, and 724 record the steps. To make a big point of officialness about one step may make it very hard to take the next step.”

  The RFC 724 team absorbed the criticism. Six months later, under Dave Crocker’s and JohnVittal’s leadership, a final revised edition of RFC 724 was published as RFC 733. This specification was intended “strictly as a definition” of what was to be passed betweenARPANEThosts. They didn’t intend to dictate the look and feel of message programs or the features they could support. Less wasrequiredthanallowed by the standard, they said, so here it was. And there it sat.

  A number of developers wrote or revised mail programs to conform with the new guidelines, but within a year of RFC 733’s publication the persistent conflict picked up again. Of particular concern, RFC 733 headers were incompatible with a mail program calledMSG(in spite of the fact that its author, JohnVittal, had helped write RFC 733).MSGwas far and away the most popular mail program on theARPANET .

  A hacker’s hacker,Vittal had written theMSGprogram in 1975 out of sheer love for the work.MSGwas never formally funded or supported, “other than by me in my spare time,” he explained. But soon,MSG had a user community of more than a thousand people, which in those days meant a huge portion of the wired world. Vittal had used Roberts’sRDmail program, which was great for handling two or three messages at a time, or even a short message stack, but Vittal was getting twenty messages a day now and wanted a program to manage them with greater ease. “WhatMSGdid was close the loop,” he said, “so that you could parcel messages out to various other files, called folders, and ultimately answer and forward.”

  Vittal, in fact, became widely known for putting the word “answer” into the lexicon of email. He invented theANSWERcommand, which made replying to messages a cinch. Recalled Vittal, “I was thinking, ‘Hey, with an answer command I don’t have to retype— or mistype!—a return address or addresses.’”

  An inspiring model,MSGspawned a whole new generation of mail systems includingMH,MM ,MS, and a heavily funded, Pentagon-sponsored project at BBN calledHERMES.MSGwas the original “killer app”—a software application that took the world by storm. Although there was never anything official about it,MSGclearly had the broadest grassroots support. It was all over the network; even ARPA’s top folks in the Pentagon used it. If anything was the most widely accepted standard, it wasMSG, which reigned for a long while. (A few people at BBN were still usingMSGin the 1990s.)

  Vittal’sMSGand hisANSWERcommand made him a legendary figure in e-mail circles. “It was because ofVittal that we all assimilated network mail into our spinal cords,” recalled Brian Reid. “When I met him years later, I remember being disappointed—as one often is when one meets a living legend—to see that he had two arms and two legs and no rocket pack on his back.”

  More than just a great hack,MSGwas the best proof to date that on theARPANETrules might get made, but they certainly didn’t prevail. Proclamations of officialness didn’t further the Net nearly so much as throwing technology out onto the Net to see what worked. And when something worked, it was adopted.

  Adventure and Quasar: The Open Net and Free Speech

  The more that people used theARPANETfor e-mail, the more relaxed they became about what they said. There were antiwar messages and, during the height of the Watergate crisis, a student on theARPANET advocated Nixon’s impeachment.

  Not only was the network expanding, it was opening wider to new uses and creating new connections among people. And that was pure Licklider. One of the most stunning examples of this began with one of the original IMP Guys—Will Crowther.

  A small circle of friends at BBN had gotten hooked on Dungeons and Dragons, an elaborate fantasy role-playing game in which one player invents a setting and populates it with monsters and puzzles, and the other players then make their way through that setting. The entire game exists only on paper and in the minds of the players.

  Dave Walden got his introduction to the game one night in 1975, when Eric Roberts, a student from a class he was teaching at Harvard, took him to a D&D session. Walden immediately rounded up a group of friends from theARPANETteam for continued sessions. Roberts created the Mirkwood Tales, an elaborate version of Dungeons and Dragons set in J. R. R. Tolkien’s Middle Earth. The game stretched on for the better part of a year and was played mostly on Walden’s living room floor. One of the regulars was Will Crowther. Where the other dozen players chose names like Zandar, Klarf, or Groan for their characters, Crowther’s was simply Willie, a stealthy thief.

  Crowther was also an ardent cave explorer. And his wife Pat had achieved renown among cavers for having been part of a small group that discovered the first known link between the Mammoth and Flint Ridge caves in Kentucky. The combined 144-mile system was the longest known cave in the world. Crowther was the cartographer for the Cave Research Foundation. He used his off-hours to plot intricate subterranean maps on a BBN computer.

  In early 1976 Will and Pat divorced. Looking for something he could do with his two children, he hit upon an idea that united Will the programmer with Willie the imaginary thief: a simplified, computer version of Dungeons and Dragons called Adventure. Although the game did not use actual maps of the Kentucky caves, Crowther based the geometry of Adventure on stark mental images of those underground chambers. The iron grate through which players passed at the start of the game was modeled on those installed by the Park Service at entrances to the Flint Ridge system. He even included a caving in-joke or two; the “Y2” inscribed on a rock at one point in the game is caver shorthand for a secondary entrance.

  Crowther finished the program over the course of three or four weekends. His kids—ages seven and five—loved it, and Crowther began showing it to friends. But the breakup of his marriage had sapped Crowther’s spirit, and he never got around to refining the game.

  Bob Taylor, now director of the Computer Science Lab at Xerox Corporation’s Palo Alto Research Center, persuaded first Severo Ornstein, then Will Crowther, to join him, and when Crowther moved to California in 1976 he left the Adventure program behind in a file on a BBN computer. Unpolished though the game was, word of Adventure had filtered through the network community.

  A Stanford graduate student named Don Woods heard about Adventure from a friend who had run across a copy on the Stanford Medical School computer, and he downloaded the game from there. But Woods had difficulty getting Adventure to run at first, and when he did he found it riddled with bugs. Still, he was hooked. “Adventure made users feel like they were interacting more with the computer,” said Woods. “It seemed to be responding more to what you typed, rather than just making its own moves like a silent opponent. I think this attracted a lot of players who might otherwise have been turned off by the idea
of playing ‘against’ a computer. This was playing ‘with’a computer.”

  The game listed Will Crowther as the author, and Woods decided to track down Crowther to get the source code so he could start making repairs to the rudimentary little program. He sent e-mail to every host on the network looking for Crowther, and finally he found him at PARC. Crowther happily handed over the code. It took several months to rework, during which the simple program doubled in size. Woods created new obstacles, added a pirate, twisted the mazes further, and added several treasures that required some problem solving before they were found.

  When Adventure was done, Woods created a guest account on the computer at the Stanford AI Lab to let people play, and swarms of guests logged in. Adventure spread like hula hoops, as people sent the program to one another over the network. Because Crowther had written it in FORTRAN, it could be adapted to many different computers with relative ease. Both Crowther and Woods encouraged programmers to pirate the game and included their e-mail addresses for anyone looking for help installing, playing, or copying the game.

  People grew bleary-eyed searching for treasure into the small hours of the morning. ”I’ve long ago lost count of the programmers who’ve told me that the experience that got them started using computers was playing Adventure,” Woods said. The game inspired hundreds of knockoffs, which eventually spawned an entire industry.

  Adventure demonstrated the appeal of an open networking culture. And the emphasis on openness grew with time. There were few closed doors on the network, and a free spirit prevailed in people’s attitudes about who could come and go through them, and for what purposes. Anyone trying to restrict the graduate student population from freely using the network would have grossly misunderstood the mind-set of the computer science community. TheARPANET was official federal government property, but network mail was being used for all manner of daily conversation.

  Then, in the spring of 1977, Quasar rolled in the door. Its arrival marked the beginning of the first debate over free speech in cyberspace. The controversy centered on an unusual device made by Quasar Industries and blew up into an argument over using the taxpayerfundedARPANETto speak, in openly critical terms, about a private company.

  The brainchild of Quasar Industries, the device stood five feet four inches and weighed two hundred forty pounds. It was called the Domestic Android robot, a programmable helper that could perform a dozen basic household tasks such as mopping the floor, mowing the lawn, washing dishes, and serving cocktails. It came equipped with a personality and speech, so that it could “interact in any human situation.” It could “teach the kids French” and “continue teaching them, while they sleep.” At the advertised price of $4,000, the thing seemed a steal.

  Phil Karlton of Carnegie-Mellon was the first to alert the Msg-Group, on May 26, 1977. His site on theARPANETwas heavily involved in exploring artificial intelligence, speech recognition, and related research problems, so he knew a thing or two about robots. The android and its inventor had attracted a fair amount of national press attention, most of it favorable. Quasar’s sales pitch had also caught the attention ofConsumer Reports,which ran a skeptical item on it in the June issue, just out.

  At first Quasar seemed nothing but an amusing diversion from the MsgGroup’s main business. Everyone in the group knew the thing was a hoax, and for a while that seemed enough. But then a sense of civic duty arose. Dave Farber told of being in Boca Raton, Florida, and hearing on the radio that the Dade County police department was considering purchasing a Quasar guard robot for their county jail, for $7,000. In March theBoston Globeran a story quoting MIT’s Marvin Minsky and other skeptical AI experts. But the article took the overall attitude, said a MsgGroup member, that it “just goes to show you, those academicians can’t do anything practical, and all you need is some guy working in the back of a garage to put them to shame.” The saga left a trail of disbelief in the artificial intelligence research community.

  Brian Reid and a colleague, Mark Fox, from the Carnegie-Mellon Artificial Intelligence Lab, posted an offbeat report to everyone in the MsgGroup, giving them a personal account of their inspection of the domestic robot, “Sam Strugglegear,” at a large department store in downtown Pittsburgh. People in the research community, knowing of CMU’s pioneering AI work, had been calling the Lab to ask how it was possible for Quasar’s robot to be so much better at speech recognition than anything CMU had produced. Rising to the challenge, a four-member team from CMU had done the fieldwork.

  “They found a frightening sight,” reported Reid and Fox. In the men’s department, among the three-piece suits, was a five-feet-two-inch “aerosol can on wheels, talking animatedly” to a crowd. Electric motors and a system of gears moved the device’s arms. The robot seemed conversant on any subject, recognized the physical features of customers, and moved freely in any direction. The crowd was charmed.

  But the scientists were skeptical. They looked around for some evidence of a remote controller. “Lo and behold, about ten feet from the robot, standing in the crowd, we found a man in a blue suit with his hand held contemplatively to his mouth like Aristotle contemplating the bust of Homer in the famous Rembrandt painting.” Reid and the others watched for awhile and noticed that whenever the robot was talking, so was the man in the blue suit—muttering into his hand. The man had a wire dangling suspiciously from his waist.

  The discussion about the Quasar robot continued on and off for a couple of years until in early 1979, Einar Stefferud, the MsgGroup’s moderator, and Dave Farber, who had been lurking on the sidelines of the commentary, sent a note of caution to the MsgGroup. “We are asking for potential problems,” they warned, “when we criticize the Quasar robot.” Using U.S. Government facilities to cast aspersions on a corporation, they said, could backfire on the ARPA research community. They urged their peers to impose careful self-censorship, to report only facts of technical interest to the community. Not everyone agreed, and with that the MsgGroup got embroiled in a soul-searching exchange.

  John McCarthy, who worked at Stanford’s Artificial Intelligence Lab, was among those most offended by Quasar’s claims. He told the group that he would not be deterred by speculation that Quasar might sue. “I think someone seems to be frightened of his shadow,” McCarthy said. “It has never been the custom of carnival snake-oil salesmen to sue their critics.” Minsky and Reid also made it clear that they would tell any reporter who asked that they believed the robot was a joke, and they’d already expressed that opinion to more than a dozen journalists.

  “I have no fear of being sued,” replied Farber. “However, we are using a public vehicle called theARPANET. We thereby expose ARPA, DOD, and our future access and use of the network to certain dangers when we use that vehicle for potentially libelous material.” Farber again urged restraint.

  Reid chimed in, saying, “[the] MsgGroup is the closest that we have to a nationwide computer science community forum.” Reid had begun to notice that the Message Group was like a social club. They had argued with each other so much that they had become friends. To restrict discussion would be unnatural. Besides, Reid took a more liberal view of free speech, reasoning that the experiment in communications would suffer if topics were restricted. “Until such time as people start suggesting the overthrow of our government,” he said, “I don’t think any sensible topic should be off limits.”

  Someone suggested attaching a disclaimer to personal communications on theARPANETso that personal opinions wouldn’t be mistaken for official business. Admitted someone else, “Who hasn’t used Net mail for personal communication? Who hasn’t spent time playing some new game over the Net? Be honest.” The passion in defense of free speech was matched by an equally strong will to self-protection; the way to protect the network itself was not to attract unwanted supervision by the government. After a few days the argument wore itself out without resolution and the MsgGroup carried on with business as usual.

  What emerged from the debate was strong evidence
that the networking community felt a deep stake in the creation of the Net, ARPA funding or no ARPA funding, and was trying jealously to guard its right to determine its future. In a realm where, in a sense, personal identity is defined entirely by the words people choose, free speech seemed second only to concern for the survival of the realm itself.

  Copper Umbilicals

  For the first quarter of 1976, traffic reports showed that the volume ofARPANETmail, compared to the volume of regular U.S. mail, was a mere ant trail in the tracks of an elephant herd. MIT’s Artificial Intelligence Lab, for example, passed some 9,925 messages during the period. (By 1996, by comparison, some sites were processing 150,000 e-mail messages every day.) MIT was a typical site, and by extrapolation, if one machine processed about a hundred pieces of e-mail a day, multiplied by a factor of 98 or so (the number of hosts then on the Net) electronic mail didn’t yet appear to pose a threat to the U.S. postal system. The post office handled more than 50 billion pieces of firstclass mail a year. But e-mail’s steep growth curve wasn’t going unnoticed.

  In the private sector, companies were poised for the concept of electronic-mail service to take off. The Computer Corporation of America soon began selling one of the first commercially available e-mail software packages, a $40,000 product calledCOMET, designed for the PDP-11 minicomputer. Another program calledMESSENGER, developed for IBM 360 and 370 computers, was soon available from a company called On-Line Software International, for $18,000. Costs were heading down, and some analysts projected a “devastating” impact on the U.S. Postal Service’s first-class business.

  “We are being bypassed technologically,” reported an assistant U.S. postmaster general at the beginning of 1976. The new technology’s growth trend and obvious potential were indeed quite dramatic. A few versions of the more sophisticatedARPANETmail programs such asMSG,HERMES , and SRI’sNLS JOURNAL MAIL, were coming into the hands of nonresearchers. Several large organizations including the U.S. Geological Survey, Department of Commerce, National Security Agency, and Gulf Oil had all started using e-mail over local area networks.