This conference was joint hosted by ACM SIGGRAPH and the Web3D Consortium, part of the greater W3C. This is the 5^th symposium devoted to research and development in the Web-based 3D graphics community. There was a special emphasis on VRML – Virtual Reality Modeling Language, a featured tool for Web3D, and Java. The conference was well and enthusiastically attended, and went a long way to sanctify this growing field as part of legitimate IT mainstream. I attended 3 workshops, which covered VRML-based art, hypermedia data issues, and GEO/VRML, an integration of GIS and VRML.

1. VRML-BASED ART

VRML art is part of the field of "telematic" art. Telematic art is interactive. Unlike traditional art, with a "passive" and "separated" viewer, telematic art is user-driven. Where traditional art is "finished" telematic art is dynamic, changed by the viewer; it is becoming to traditional art’s being. VRML is the favorite tool of telematic artists. The user can fly through it, surround it, go through doors to multiple layers of it, etc. It is "immersive".

This was a good topic for an educator to begin with since it prompted many pedagogical speculations about the value of interactivity. Telematic art is informal and subjective. As a viewer, you are invited to assert your own persona into it. In fact, that is a common theme of all hypermedia driven art – including film. You can dictate events and scenes. This continues a trend from Picasso, Braque, DuChamp and others – inviting the viewer to so impose their own feelings, thoughts or imagination into a scene as to be master of it. Like God, the subject has an immediate 3D and 4D experience of their own cosmos. But it is not the pure essence or being we associate with Plato or Aristotle’s Unmoved Mover". This is more like Heraclitus’ god of becoming. Below are three examples by Tania Fraga, a Brazilian telematic artist who hosted this session of the conference.

Immersive art – subject can wander along any trail, change directions and jump to a new trail

"Poetics of Becoming" - follow trails that are both image and text – a good example of the hybrid nature of telematic art

How well this works as art is, I suppose, a topic of some debate. Certainly the potential for learning technology is apparent. VRML-driven environments can integrate many of the most acclaimed tools of learning technology and channel them powerfully, but to what script? This is where the debate beyond the technology lies. Telematic artists are driven by various theories, such as Gestalt or Calvino’s "6 proposals for the new millenium", which provide prescriptions for what is appealing aesthetically, but these are not exactly learning theory. How this technology could be used to build out learning precepts is relatively undetermined at this point – a field waiting for some research.

2. THE DATA PROBLEM - HYPERMEDIA

Hypermedia data is heterogeneous objects that enjoy temporary relationships to one another, have temporal and spatial dimensions, and have a user and context dependence that can be very subjective. This session concentrated on hypermedia, particularly VRML, data modeling issues.

Modeling this data offers some serious challenges. Unlike traditional data, it has no a priori standardized feature set, and therefore nothing obvious to query on. For example, below are 2 VRML objects in the same category car/model, but one is a hardtop and the other a convertible

How can we build a query that will capture both, but make the distinction between the two types? We could use descriptors, as with traditional database retrieval, but what if we wanted to use features instead? And what if we had these models?

Or these, where we are comparing a 3D object with a 2D image:

Now suppose beyond these simple queries you want to do more powerful things, like piece together a new object with parts of others, or have objects interact in what if scenarios, e.g. crash tests? Better mechanisms are needed for modeling spatial features, behavior, and event/interaction. These capabilities need to be built into a browser that can capture user profiles and preferences. Users also need metrics that assure them that they are getting quality results from their queries and simulations.

VRML produces the objects this data model needs to capture. VRML is built on a node structure, in which each node has its own atomic role, but can be dynamically structured into a hierarchy to build complex objects or events. These nodes are inherently temporal and spatial. You can synchronize events, and objects can interact. VRML describes a real world – it is a conceptual data modeling language. What data model can handle its features? Relational and EER cannot represent behaviors. OO has a rich semantics but cannot capture dynamic structures, and while it can capture spatial data it cannot model with it, e.g. build a spatial hierarchy.

Using XML to model hypermedia data has some potential. XML extends metadata to text and object components, giving users more context control. XML linking and query languages add additional capability for handling interaction between spatial objects, multidirectional links for example, and the ability to replace content. This last point has special implications for VRML objects, since it allows users to alter these objects or their component parts - replace a door on a car for example - rather than just traverse them.

While XML may work as a logical data model, but it is not useful as a conceptual data model (does not describe the real world), nor as a physical data model (says nothing about a storage scheme). It could provide a standardized grammar for VRML worlds, but lacks proper "modeling" functions – like behavior, event interaction, and spatial modeling features. There are other hypermedia languages available – HyTime, MHEG, and SMIL are examples - that can handle event-driven synchronization, and dynamic associations between atomic nodes. However, like XML, these languages do not deal with data storage issues – prefetch, caching, etc. and they do not return any metrics. UML is another possibility, but the presenter did not seem to know much about it. The fact is, that hypermedia data modeling is an area that needs more research, and remains an obstacle to Web3D development because of this lacking.

GEO/VRML

At the Monterey conference, the group delivered nine new VRML nodes that support geographic coordinate systems, integrate data sets from different sources and coordinate systems into a single global context, and perform geographic-based animations.

These nodes represent some extremely important progress. GIS-based data modeling is a seasoned and well-developed practice that is in increasingly hot demand. To marry that technology with the power of VRML offers huge opportunity for 3D and 4D modeling. It is also viewable and interactive via Web browsers.

GEO/VRML model of Yosemite, view from SW

Hurricane Linda from NOAA GOES-9 satellite, color texture map created by combining visible and infrared bands

VRML capability comes as plug-ins. In this session it was recognized that these tools would better serve Web3D’s goals if they were native code in browsers. Some start-up companies are working on that. Another group, SEDRIS, Synthetic Environments Data Representation Information Systems (see http://www.sedris.org), also a Web3D working group, is working on amalgamating these GEO/VRML worlds with physical science models from hydrology, natural resource management, and other fields.

CONCLUSIONS

There is great promise for future work, and the spread of these tools will surely also encourage further development in developing better data models for VRML and hypermedia. The vision is that someday these hybrid technologies will be wedded to a fully developed component-based data model. Then users will be able to build and dis-aggregate new constructs at will.