To create a networked VRML2 world right now you must use CGI

• Create libraries that allow you generate modular solutions.
• Output custom represenations from user input "on the fly".
• Generate transactions via URLs.
• By inlining objects which have the ability to interact
  with CGI, a dynamic interface is possible.
• Many applications possible.
• Allows developers to prototype before integrating
  more robust solutions.

When a robust networking solution is available via a plugin API
real hypermedia applications will be possible:

• Multiuser Worlds
• Intellegent Search Robots
• Informational Spaces.

It's not near, its here.