• Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint
Share this Page URL



Mark Pesce Co-inventor, VRML

Worlds in the Web[*]

[*] © 2000, Mark Pesce. All rights reserved.

Before the Web, when virtual reality had barely escaped from the lab to make its way into Japanese department stores, real-time 3D meant half-million-dollar Silicon Graphics workstations running industrial-strength operating systems. Heady times—but only for a lucky few.

I wasn't one of them.

I read the articles—in Scientific American, the Wall Street Journal, and Mondo 2000—and got the concept immediately. Virtual reality meant the visualization of the imagination, an opportunity to take a peek at things that had only ever been touched-in poetry or art-but never truly realized, never made tangible.

After reading William Gibson's Neuromancer I found myself energized by his vision of cyberspace—the "consensual hallucination" which would become the communications playground of the twenty-first century. Working as a software engineer at Shiva, I knew something about communications; I was convinced that the networks that Gibson envisioned would soon spread ubiquitously across the world. But networking in 1990 felt more like a string between tin cans compared to the rich universe of Gibson's cyberspace. The networks were coming into place, but the interface to these networks remained light-years behind, an anachronistic opening to an undiscovered world.

Over the next few months, absolutely consumed, I began to think about how to tie the ever-expanding world of the Internet (which I had been working on for several years) into the imaginary worlds of virtual reality. My work at Shiva gave me an enormous background in networks, and my reading taught me lots about VR. On March 7, 1991, I had an insight which forever tied the two together in my mind.

In Neuromancer, Gibson wrote about the network and its visualization—cyberspace—as if they composed the twinned strands of DNA, each incomplete without the other. It seemed to me that the FTP sites which spanned the Internet—even in 1991—could be considered just as real as any of the "places" in Gibson's vision of the future, and—with a little work—it might be possible to make them as "real" as anything Gibson had described.

Shiva began as a Macintosh peripherals company, and most of the engineers used Macintoshes; when Spectrum Holobyte released Spectre, its enormously popular 3D networked video game, it became our passion; week after week we'd blast each other across the company network. Spectre ran just fine on underpowered Macs—now I knew for sure that all of those big computers and defense networks would soon be overrun by cheap machines connected through the Internet.

I could see the future—at least, I believed I could—and I couldn't think about anything else. In early 1992 I left Shiva to found Ono-Sendai (named after the company in Neuromancer that manufactured "cyberspace deck" portals into the infosphere) and set out to create an inexpensive VR system that could connect to the Internet and create a shared VR experience for game play, for education, for anything imaginable. That goal turned into a bigger task than any of us had imagined.

Ono-Sendai survived about eighteen months before it ran out of capital and collapsed. It did, however, put me in the right place, at the right time, to witness an almost insignificant event. I subscribed to a number of mailing lists, including one calling itself FRINGEWARE, dedicated to cutting-edge reports of the weird or unusual. Around June 1993 I began to see strange bits of computer effluvium show up in messages:

I had no real idea what this was, only that it was called a Uniform Resource Locator, or URL, or "earl," and that it referred to a document stored on something called the "World Wide Web." All of this meant nothing to me at the time, just an impenetrable bit of hacker-speak that I promptly ignored.

A month later I made my first visit to SIGGRAPH, the annual technical and trade show for the computer graphics industry. My friend Coco Conn had an area for SIGKIDS, bringing students from ages eight to eighteen in contact with the latest in technology. The SIGKIDS got an opportunity to play with the kinds of high-end computers they'd normally never see until they got into graduate school. While there, I walked over to a lonely Silicon Graphics workstation which seemed to display nothing more interesting than a text document. A few of the words in this document, underlined and in blue, seemed to hold some hidden meaning. I moved the mouse over them—it turned into a pointing finger—and I clicked.

And the whole world changed.

I had just traversed my first link on the World Wide Web, using a week-old browser known as NCSA Mosaic. It seemed entirely unimpressive at the time—the links went to pages with more text, and further links went to even more text—but I got the concept.

After SIGGRAPH, I returned home and loaded NSCA Mosaic onto my refurbished SPARCstation, loaded the NCSA WWW server software, logged onto the Internet, and—abracadabra!—my machine became Web site number 330 in a small but growing universe of hypertext. Every night, for two weeks, I came home from my consulting job and surfed the Web. And then I was done. I'd seen everything, gone everywhere, read every word on the Web—including a handful from a man named Tim Berners-Lee, half a world away in Geneva, who talked about future directions for the Web, including virtual reality.

It all clicked: I had bought the SPARCstation to finish my work on Internet protocols for virtual reality, but I had wondered what to put in this cyberspace once I'd finished the job. Tim gave me the answer: Make worlds in the Web. I'd gotten my marching orders. I had all the parts of the vision that would, just a few months later, become VRML.


In December I received a phone call from Tony Parisi, with whom I'd previously had an acquaintance. I didn't know that Tony was a serious hacker in his own right or that his skills perfectly complemented my own.

On New Year's Day, 1994, as they moved into their new apartment, I gave Tony and his wife, Marina, an overview of my new ideas regarding VRML and cyberspace. Tony, interested, asked if there was any way he could help. "Absolutely," I agreed. "You want to work on this 3D browser for the Web?" He had already heard of the Web and said, sure, what the heck, let's do it.

Over tall, strong coffees at our favorite coffeehouse—Jumpin' Java—we sat and plotted the design. Tony was an expert in computer languages; he knew how to translate text into commands the computer could understand and execute. I knew how to connect the computers over the Internet, and—because of a gift of some software from a tiny company in London called Rendermorphics—we had a high-performance 3D graphics package to make the 3D worlds visible. We quickly sketched out a language that would create 3D spaces inside the Web. Tony took a few days to write the language parser; I took a few more to integrate it with the Web and tack it onto the Rendermorphics software. By early February we were finished. What we'd created wasn't much to look at; you could click on a link inside of Mosaic, then another window opened a "helper application" that displayed a 3D projection of a cube, which could be rotated and examined from any angle. Labyrinth was not very impressive, but it was a beginning.

Going Public

Tony thought the whole thing would end there—what else could you do with it? But I had greater plans, and I knew someone who would care about what we'd done. I fired off an email and told the inventor of the Web that we'd realized his goal of worlds within the Web, space in cyberspace. Within a few hours Tim had answered, congratulating us on our work and inviting us to present it at a conference he was planning for the Spring, at CERN, the gigantic European atom-smasher which was birthplace to the Web.

So, in May, Tony and I pooled our resources to send me to Geneva. The First International Conference on the World Wide Web, an incredible week-long festival for hypertext hackers, brought the entire community together for the first time. On the first afternoon of the conference fifteen people got together in a tiny meeting room to discuss virtual reality on the Web.

Tim was there. And so was Dave Raggett. Dave, who had been intimately involved in the creation of HTML, proposed that the Web needed a "virtual reality markup language" or VRML—and the name stuck. (Later, "markup" was changed to "modeling," which more accurately reflects the function of the language.) I offered the source code that Tony and I had developed to the Web community so that it could become a foundation for an extended exploration of cyberspace visualized.

In retrospect, that was one of the wisest decisions we could have made. It gave all of the loose efforts toward development of VR on the Web a specific focus, and it brought like-minded folks into our orbit—people like Brian Behlendorf, a 19-year-old setting up shop as the first Webmaster for WIRED magazine even while moonlighting as a co-founder of Organic Online, the first Web design firm. Brian offered to establish a Web site and www-vrml mailing list on WIRED's servers, giving us a place to house a virtual community of VRML enthusiasts. Within a month two thousand people from around the world had joined www-vrml to discuss the requirements and possibilities of virtual reality on the Web.

In early July, Brian dragged me down to Silicon Graphics to meet up with two researchers, Rikk Carey and Gavin Bell, who had created their own 3D language, known as Inventor. Far more evolved than our own simple efforts, Inventor proved to be the foundation for VRML 1.0—so Brian is, in his own way, very much responsible for VRML.

Daniel's Story

Tony and I realized that we needed a project—a "killer app"—that could thrust VRML into public consciousness. Invited by Coco Conn to show our work at 1994's SIGKIDS, we began to cast about for a project that could do justice to our own work. We found it in the Holocaust Memorial Museum, recently opened in Washington, D.C. One of the tours through the museum—designed specifically for children—told "Daniel's Story," the sad tale of a young Jewish boy growing up in Nazi-era Berlin, then forcibly relocated to the Warsaw Ghetto, and, terminally, to Auschwitz. We used our simple VRML browser to create a model of Daniel's room in the ghetto. As in the real-world installation, Daniel's "notes"—fragments from his diary—tacked onto the walls, informed visitors of his daily struggle for survival.

With a lot of help from C. Scott Young—the first VRML modeler—and plenty of encouragement from the Museum, we finished the project just hours before SIGGRAPH opened its doors at the Orlando Convention Center. Ten hours a day, I sat and guided children through a story which many of them had only vaguely heard of. The kids got it immediately; I knew we'd struck a chord, at least with the younger set.

To my surprise, the media began to drop by. At that time, VR meant huge, expensive systems; in our modest booth we showed a full VR system running on an unadorned PC connected to the Internet. This, put simply, blew their minds. One prominent VR journalist confessed that this was the most impressive thing he'd seen at the show. A buzz began to develop.

The Big Bang:VRML 1.0 and 2.0

On Wednesday evening, SIGGRAPH attendees were invited to take part in the first VRML "birds-of-a-feather," where we got together to brainstorm about future directions for VRML. Rikk Carey put forward an exciting proposition: why not use Silicon Graphics' Inventor to make VRML industrial-strength? After a handshake deal, Gavin Bell began to work with Tony and me on the design of the VRML 1.0 specification, which we presented at the Second International Conference on the World Wide Web, just two months later!

The first versions of VRML were designed to emulate the Web of 1994. The only "interactivity" in the early Web came from clicking on links and going to new pages or VRML worlds. If this seems rather simplistic, it can be traced to the fact that HTML—the content of the Web—expressed the "look" of things, but not how they'd "play." To make the Web playful, you'd need a standard programming language, which could be written once and run anywhere, on any computer or operating system. Such a language did not exist in the world of 1994.

It was back to the drawing board for us; just after SIGGRAPH '95—where we saw VRML explode across the trade show floor—we secreted ourselves away with Rikk Carey, Gavin Bell, and a few others, a hand-picked crew with a mandate to create a next-generation VRML.

We didn't succeed. Too many good ideas competed for their place in the specification. What about JavaScript? What about Java? What about sound? What about event models? We decided that engineering by committee was a bad idea, so we threw the problem back to the VRML community, creating a set of recommendations and asking for proposals.

By January 1996, we'd received a dozen different specifications from companies like IBM, Microsoft, Sony, and Silicon Graphics. Each of them had merits, and each had shortcomings. We reviewed them, made some comments, then left it to the community to decide which proposal would become VRML 2.0. The Sony/Silicon Graphics proposal—"Moving Worlds"—which successfully integrated Java, JavaScript, and a sophisticated event model, became the basis for VRML 2.0. In May, VRML 2.0 browsers began to appear.

It was like the Big Bang; a legion of content developers began to show off their own imaginings, shared over the Web. Although by no means perfect, VRML 2.0 provided a platform for a new kind of network visualization. It had taken a few years, but with companies like IBM and Microsoft promoting it, VRML now seemed poised to conquer the world.

VRML97, X3D, Java 3D, MPEG-4, and Beyond

By early 1998 it became clear that VRML, for all of its power, had some serious drawbacks. The VRML browsers required multimegabyte downloads, a big turn-off for Web surfers on slow modems; they ran slowly, at least for users conditioned to the lightning-fast worlds of Quake or Tomb Raider; and they didn't integrate very well with the Web environment.

Microsoft, which had sensed that 3D for the Web would be a very powerful tool for presentations and visualization, began work on a project called Chrome. Chrome differed from VRML in one very significant way; rather than conforming to the "scene-graph" structure of Inventor-like languages (like VRML 2.0), it belonged to the family of XML languages—giving it a closer relationship to the Web browser. Chrome looked a lot like HTML and could be dropped directly into a Web page, interacting with all of the elements on that page. Soon, it seemed, 3D would become an integral part of the Web environment.

Microsoft backed away from Chrome soon after they released it, but the idea had caught on, and the many companies involved in VRML work began to sketch an architecture they named "X3D," a combination of XML and VRML which would take the best ideas in Chrome and marry them to the existing base of VRML applications and content. In the year 2000 the wraps will be taken off X3D, and-finally-virtual reality will become synonymous with the Web.

In 1997 VRML 2.0 evolved into VRML97, an International Standards Organization (ISO) specification that guaranteed a worldwide standard for Web3D and pioneered the way for other forms of Web3D. Java 3D borrows heavily from VRML, MPEG-4/BIFS is directly based on VRML97, and X3D is the official successor to VRML97, due out in 2000.

It makes little difference which of these becomes the standard for Web3D; each may become a de facto standard—four standards for four universes. VRML works well as an interchange format for real-time 3D and is really designed as a modeling tool; Java 3D provides the low-level system control that graphics hackers demand; MPEG-4 delivers 3D over a variety of network connections and devices, from the Web to broadcast television; and X3D allows Web designers to drop 3D into their pages without concern for the plug-in architecture or client platform. These technologies complement each other far more than they compete, and none of them is likely to go away anytime soon.

Inside this book you'll find the technical treatises which underlie each of these different approaches toward Web3D. Sometimes they contradict, sometimes they run in perfect concordance; but, differences aside, each can give you the ability to articulate your own feelings, your own dreams, your own imagination. And that, in essence, is what this book is about, what Aaron Walsh and Mikaël Bourges-Sévenier have worked hard to bring you. Learn, and you'll speak a new language—unlike English or German or Japanese—but just as powerful in its ability to express what we are.

And that makes it something worth learning.

Visit http://www. CoreWeb3D.com/go/forward/ to read an expanded version of this document, which includes Mark Pesce's view of the future of Web3D content.

  • Creative Edge
  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint