In an informational space, there is one overriding consideration: how much "space" (in terms of bytes of storage in the real world; an approximation of the Washington Monument may require less storage space than the structure of a complex protein molecule) a given object takes up. The more dense, the more "realistic" (if the added information is used to simulate "realistic" detail) a given object is, the more space is required to store it, and the more time is required to transmit it.
* INFODUMP * INFODUMP * INFODUMP *
First off, you've got to understand that there's this thing called bandwidth. No matter how cool the Net architecture is--and trust me, Net architecture is a way cool topic that can put most people to sleep faster than you can say "Transmission Control Protocol with Internet Protocol"--the amount of data that can be pushed through the Net in real time depends on how "wide" the data channel is. To do Net virtual reality in real time you need, at minimum, a bandwidth as wide as a TV channel. That's wide.
--Bruce Bethke, Headcrash
Virtual reality, in its present form, involves trade-offs. Transmitting video, for example, is usually infeasible because one cannot transmit the number of images per second needed to produce the illusion of motion. Work-arounds are required, and several "solutions" have been devised: compression of video data (so that each frame occupies less space, or only the relevant parts of a frame are updated, or both.); transmission of parameters necessary for the user's machine to recreate the frames; or the omission of unimportant data (the real reason why games like DOOM move so fluidly).