The business press is fascinated by any hint of virtual reality on the Web, and there's always a fuss when someone announces the latest proposal for the One True Standard that will make VR real (and make them rich beyond their dreams). If graphics are good, immersive, virtual graphics will be better. VR, they assume, will transcend or replace hypertext.
People aren't particularly good at working in three dimension. Our bodies and senses are adapted to plains and grasslands; we are adaptable ground-dwellers who can, on occasion, do things in three dimensions.
People are good at kinesthesia -- at knowing where their bodies are situated and how the edges of their bodies relate to nearby objects. Kinesthesia lets you drive your car very fast, only a few inches from a bridge abutment or another automobile -- but it still doesn't help you know, while you do it, where in the world you are.
We are visual creatures who perceive two dimensions well and who see the third dimensions poorly. While people inhabit three dimensions, the often describe the way they move in two; when people build three-dimensional living spaces, they usually flatten each unit into a simple, approximately two-dimensional topology.
Three dimensions aren't enough; hypertext structures don't map much better into three-dimensions than into two dimensions. It is true, of course, that the familiar box-and-line diagrams we sometimes use to represent hypertext structures are easier to draw in three dimensions than in two; in three dimensions, the lines don't keep crossing each other. But the connections between complex ideas are not well represented by simple geometries or concrete objects. Well thought-out links can express relationships that mapping everything onto virtual walls, rooms, cars, and dragons cannot.
Three dimensions are too many; mapping hypertexts into space forces us to accept juxtapositions that are irrelevant or distracting. This is a problem well known to museum designers: paintings in a room influence each other in ways that their creators did not foresee. Juxtaposition can express critical ideas, but it can also be merely distracting or inadvertently comic.
Instead of building literal simulacra of three-dimensional spaces, though, the hypertext writer is free to build artificial spaces where juxtaposition is controlled, where things that should be linked are connected, and things that should not be linked are not accidentally pushed together. We create links to make connections where connections make sense; we need not link unrelated things simply because they occupy the same virtual room.