VRML, THE VIRTUAL REALITY MARKUP LANGUAGE
One of the popular topics at the recent World Wide Web conference in Geneva
was how to implement distributed virtual reality on the Web. Mark Pesce and
Brian Behlendorf, who didn't know each other before the conference, want to
harness that interest to accelerate the development and acceptance of an open,
platform-independent standard for Web-based VR. To that end, they have started a
mailing list(6) for the Virtual Reality Markup Language ( VRML) (7) and put some
information about VRML and the list participants on the World Wide Web.
Early discussion on the VRML list ranged over topics as broad as the
questions in the introduction to this issue of Release 1.0. Behlendorf and Pesce
have kept the list focused so it can have concrete results, which others can
then experiment with more broadly. In that spirit, the VRML list participants
are now reviewing 3D scene- and object-description languages. Candidates
include:
* CDF, the Cyberspace Description Format, developed by Autodesk.
* FFIVW, the File Format for the Interchange of Virtual Worlds, by Bernie
Roehl and Kerry Bonin, is from the computer-games industry. The name speaks for
itself.
* Meme, Immersive System's Multitasking Extensible Messaging Environment,
proposed by Marc de Groot, president of Immersive, and described in Release 1.0,
6-94.
* MSDL, the Manchester (UK) Scene Description Language, derived from the
next candidate, OOGL.
* OOGL, the Object-Oriented Graphics Language, developed at the Geometry
Center in Minnesota and popular as a geometric visualization format in
mathematics, physical sciences and engineering.
* SGI's OpenInventor, built atop OpenGL, SGI's 2D and 3D graphics library.
This standard has a lot of history and support; it is the front-runner.
* TSIPP, an extension of Tcl (the Tool Control Language, see Release 1.0,
2-94), developed by John Ellson.
The participants will vote soon; then some of them will develop browsers,
compilers and authoring tools that conform to the newly anointed VRML 1.0
specification. Later, they expect to offer the specification to the recently
formed World Wide Web Consortium, which includes MIT, CERN (where the Web was
created) and the National Center for Supercomputing Applications (where Mosaic
RELease 1.0, August 3, 1994
was developed). By that time, VRML will have been through considerable peer
review, which should speed acceptance. Eventually the members of the VRML list
may present it to the Internet Engineering Task Force, which sets "official"
Internet standards.
The VRML 1.0 spec will not support real-time interaction between multiple
avatars, but will focus on making possible a concatenated "space" designed and
shared by many individuals. In principle, object attributes could change
frequently (e.g., through database lookups at runtime), though it is not easy to
have them change once they are loaded. The designers of VRML 1.0 are leaving
room to include interactivity in 2.0.
Behlendorf has plenty of experience moderating lists: He started the SFRaves
mailing list several years ago. He's an undergraduate student in Computer
Science at UC Berkeley and Wired Magazine's Webmaster. At Wired, he's helping to
create HotWired, the Magazine's online presence, which will likely use VRML
when it's available.
Pesce and the Labyrinth Group
At the Web conference, Pesce, president of the Labyrinth Group, described
his proof-of-concept scene-description language, also called Labyrinth. The
RELease 1.0, August 3, 1994
first two paragraphs of this issue are an embellished example of how Labyrinth
could assemble and render scenes. Designers use it to define objects, scenes and
camera positions, as well as other elements such as texture maps, lights and
sounds. Each element might be stored on a different server; they are assembled
dynamically when the scene is invoked.
Pesce displayed Labyrinth integrated with the World Wide Web at SIGGRAPH
this month, in a special section called SIGKIDS. In collaboration with the US
Holocaust Memorial Museum and Husky Labs (which manages the Museum's Web site),
he created a virtual exhibit called Daniel's Story. The virtual exhibit
recreates some of the rooms in the Holocaust Memorial. "Visitors" can move
around in the rooms, walk around the objects in them (beds, desks, chairs, rugs,
paintings) and read Daniel's diary pages, which describe the setting and the
events.(8)
When visitors select a diary page, for example, they travel across the Web
to typical Web pages. Conversely, when someone first invokes Daniel's Story from
a Mosaic Web browser, MIME(9) automatically launches Labyrinth, which assembles
the files, parses their contents and presents them. That makes it easy to move
from Webspace to cyberspace and back.
RELease 1.0, August 3, 1994
Origins, destinations and the Cyberspace Protocol
Pesce created Labyrinth more to get people charged up about the
possibilities of VR on the Web than to become the universal standard for VR on
the Internet. He expects one of the submissions listed above to become the core
of VRML 1.0, most likely OpenInventor, and will support whatever wins with
public-domain browsers by the end of 1994. Now he funds operations from
short-term projects. Interest in distributed virtual environments is heating up;
Pesce is in discussion with companies about developing information services,
training environments, commercial spaces and games.
Pesce has spent 15 years in networking and data communications, first as an
engineer at Shiva, where he developed dial-up networking systems for Macs and
PCs. William Gibson's novel Neuromancer triggered Pesce's longstanding interest
in virtual environments and galvanized him into action. In 1991, Pesce started a
company called Ono-Sendai to develop the software and hardware necessary for
widely distributed virtual worlds (yes, Ono-Sendai is named after the vendor
that makes the "deck" with which the protagonist in Neuromancer "jacks in" to
cyberspace). He left Ono-Sendai in 1993 to start the Labyrinth Group.
Pesce believes that as more stuff is put into cyberspace, people will need
some way to map and agree on a unified space -- a universal coordinate system
RELease 1.0, August 3, 1994
that spans the entire Internet and provides conventions of scale and place.
Gopher-style hierarchies and Web-style link traversal aren't helpful enough:
Users get lost quickly and don't have enough meaningful context or framework.
He believes the underlying representation should be Cartesian, with some
added features. Pesce has created a method for addressing and navigating virtual
spaces called the Cyberspace Protocol, which he wants to offer to the Web and
broader Internet communities. People can have their own mappings, says Pesce,
but somewhere there should be a single 3D space with agreed-upon coordinates.
(3)We introduced LambdaMOO briefly in Release 1.0, 6-93 and 6-94. MOOs are
textual, multi-player environments that allow participants to modify or create
objects, characters and places. For a further description of MOOs and MUDs, see
those issues, as well as 7-93. MUDs are multi-user dungeons; MOOs are MUDs that
are object-oriented.
(4)Jupiter users have microphones and (optional) cameras mounted near their
computers. The video refreshes at only five frames per second, has medium
resolution and doesn't synchronize with the audio feed, but it is good enough to
show whether others are present, have visitors or are absorbed in activity.
Besides, several of these videos can be running in one palette at the same time
(using the Internet multicast protocols) -- all on standard Ethernet.
RELease 1.0, August 3, 1994
(5)The usage of the terms "communities of practice," "periphery" and
"legitimate peripheral participation" at PARC derives from studies of learning
by Jean Lave and Etienne Wenger in 1991 (see Resources, page 18).
(6)Mailing lists are highly democratic discussion tools because they can
include users with the lowest possible Internet connectivity: an e-mail gateway
to the Internet. Any message sent to the list address is reflected to all
participants, who can subscribe or unsubscribe as they see fit by sending
messages to a separate administrative address.
(7)The name VRML, coined by Dave Raggett of HP Labs in Bristol, UK,
reflects its ties to HTML, the Web's HyperText Markup Language.
(8)To view the starting page, point your graphical Web browser (e.g.,
Mosaic, MacWeb) to <http://www.butterfly.net/lab/daniel/sign.html>.
(9)For more on MIME, the Multipurpose Internet Messaging Extensions, see
Release 1.0, 2-94. MIME intercepts and dispatches requests for viewers,
interpreters or for further processing that may accompany e-mail message or Web
network requests.