Browsing today
Navigation through information available on the world wide web nowadays is a one-dimensional task: The whole world of information spans between the "backward"- and "forward"-button of your favorite browser. "Going" back to a piece of data you spoted a few minutes ago forces you to predict the number of times you would have to click on one of these buttons in order to get there. Other concepts as history-logs or bookmarks that aim at supporting you in digging up sufficient answers from your memory to daily questions like "where did I read this article on ... again?" or "I just saw this cool gadget on the net a couple of days ago, now how can I get back there?" also happen in a very flat world: Scanning endless lists of text. Sometimes, as CGI, PHP and others get into game these lists really become a mess: What was the article about that I bookmarked last week under "http://www.wired.com/ news/technology/0,1282,36328,00.html"?

"Surfing the net"? There is no such thing!

webray: three-dimensional information architecture
The idea behind Webray is to use one of the most fundamental abilities of human beings: Spatial knowledge, knowing your own location in the world at any given moment. What our ancestors developed in order to get the food back to the cave, Webray takes advantage of to navigate through the parallel world of information on the net.

We have developed a prototype that represents HTML-pages as three-dimensional geometries. Additional cues as the title of the page and a picture taken from it are mapped onto the shape to give every one of the generated objects a unique and identifyable appearance, even looked at from a distant point of view. Clicking on one of these objects for 20 msecs brings up a standard browser window that displays the page symbolized by the item. If the user shows further interest in a certain object and moves (flies) towards it, the object would unfold, build a horizontal platform and smaller representations of the links contained on the page would be arranged on it. These new objects are scaled down to keep the dimensions of the whole environment consistent. This action can be repeated again and again, leading to a multilevel self-similar landscape as in a fractal or a Lindenmayer-system. To keep the travelling speed relative to these smaller objects the same as on lower levels, the user and his or her "footsteps" are scaled down as well. Moving away from a higher-level-platform to a lower-level-platform again, rescaling happens seamlessy to match travelling speed and observation-height again.

By doing this, we enable users to move from one site to another allways knowing they just would have to turn to their left and move a couple of blocks on to get back to where they where just moments ago. Using the system for an hour results in a complex, individual information landscape, that is still navigatable in a very intuitive way. To provide further help, we integrated additional possiblities to costumize these landscapes: Users can delete single objects they considered uninteresting or rearrange whole platforms (i.e. to make up groups of sites related to certain topics.)

To provide a starting point we inserted a search engine-interface that lets you do standard searches, even including boolean operators. Results then are generated in the described way, ready to be explored.

Technology behind webray
Technically, this early prototype of Webray is based on a combination of a VRML-browser-plugin and a java applet interfacing via EAI. For reasons of performance and stability we use the blaxxun-plugin. The searchengine interface is connected to google.com. Results coming back to the searchengine are parsed, bitmap-textures of the titles are generated on-the-fly (as VRML-text would be too slow), a picture is mapped and individual threads are created to do further preparsings of pages linked within the original document. Until now, one of the main problems is to get out of the java-sandbox to enable the applet to do http-requests. Unfortunatly at the time being we have to ask users to switch of some security-features of the hosting browser. To implement this first prototype quickly we also used a free scenegraph API for VRML called JVerge, which leads to some overhead. This will be stripped of in future versions.

Future development will concentrate - besides perfomance issues - on the exploration of additional techniques of arranging information in an VE. Being students of architecture we will have a close look on techniques originating from urban planning, inner organization of rooms and aesthetic issues. Also, related research we have done in the fields of human-computer-interaction and creation of immersive virtual environments will be applied , i.e. in evaluation phases.

We dream of a truly threedimensional information space, making movement in it as natural as walking through a city, without the hassle of sour muscles.