5.8 Beyond HTML

Ever since the publication of Vernor Vinge's groundbreaking science-fiction novelette True Names in 1978, many computer programmers and users have been fascinated by the ideas of ``virtual reality'' and ``cyberspace''--computer interfaces using 3D surround graphics with which a user could interact in something like the way our physical bodies interact with the real world.

In 1994, two Silicon Valley programmers with backgrounds in computer graphics went public with the provocative idea that the distributed-hypertext model of HTML might be used as a foundation for building cyberspace. A working group began adapting and enhancing a graphics-description language previously developed at Silicon Graphics Inc. to be used with HTML and Web software.

The result was Virtual Reality Markup Language (VRML). VRML follows the HTML model in several important respects. Like HTML, VRML is a content-based, plain-ASCII markup language that describes its universe at a high level, and leaves it up to the browser or client program to make detailed presentation decisions. VRML technology has since evolved even more rapidly than the Web itself. As of January 1996, several first-cut VRML browsers are available, including an add-on for the popular Netscape browser for PCs.

VRML documents describe worlds populated with 3D shapes. Developers build objects from basic shapes, such as cubes, cones, and spheres, with a variety of surface effects, including texture maps and lighting, available for composing realistic objects. It's possible to associate URLs with objects in a VRML world in such a way that when a user touches the object, they are transported to another VRML world, an HTML document, or a CGI script (which itself may generate a VRML world).

Some intriguing demonstration interfaces to large databases have already been built and suggest the beginnings of true cyberspace architecture. So far, however, these demonstrations fall short of the strong virtual-reality experience depicted in science-fiction books and movies. Three related problems bar the way. The design problem is that VRML currently has no capacity to do animation. VRML worlds are static. While it's easy to move the user's viewpoint in a VRML world, the only way to actually move or alter a VRML object is to generate an entirely new document and render it. Proposals to support animation description are in the works, but none have yet been adopted as of January 1996. This may well have changed by the time you read this.

The second, practical problem is that reasonably priced desktop machines in 1996 simply don't have the capacity to do realistic animation of general scenes. The computational cost of point-plotting, hidden-line elimination, texture-mapping, and lighting calculations gets very large, very quickly. Companies that care about animation, like Industrial Light and Magic, use multimillion-dollar special-purpose supercomputers for this job because they have to. Current VRML browsers, for example, frequently have to drop back from full-surface rendering to wire-frame mode when moving the viewer's eyepoint. Even without animation, rendering of static images including texture-mapping, spheres and sculpted surfaces can take many seconds, which is unacceptably slow.

The third problem is that VRML worlds are not yet shared spaces. For virtual reality to begin to approach the interactive richness of the real world, it must be possible to interact in a VRML world not merely with objects but with the ``avatar'' or cyberspatial manifestations of other users. This is an area of very active interest to the VRML developers (many of whom have been directly inspired by the Black Sun environment depicted in Neil Stephenson's immensely popular science-fiction novel, Snow Crash), but no single solution has yet emerged.

(There are other, comparatively minor problems: VRML has no sound capability, for example. But it's easy to imagine a solution for this one within current technology.)

Thus, VRML is still currently more a promise and a technology direction than a complete answer to the virtual-reality question. But the trend-curves of desktop hardware performance and the amount of money being poured into better multimedia graphics accelerators make it one worth careful watching. With another few years' hardware evolution and a few foreseeable VRML language enhancements, a true distributed cyberspace, built around and using WWW servers, may well be within our reach.