What I learned at SigGraph98

Java and JavaScript Interface to VRML:

I attended the last part of the VRML course to see what they had to say about controlling VRML with different scripting languages. Go Here for my Complete Course Notes

• VRML2.0 currently supports several different scripting languages including: Java, JavaScript, VRMLScript, and ECMAScript... As far as I can tell all the *.script languages are basically the same
• For simple and easy tasks JavaScript will be what you want to use since it is less complex and easier to use than Java... however more complicated things will require the Java interface
• Java provides access to the network more complex math, Data Structures, classes and sockets are provided in Java and not in JavaScript
• With Scripting languages is it possible to Add and Delete Routes from the VRML dynamically. You can replace the whole world dynamically with Browser access...
• Scripting also makes it possible to subclass the different VRML nodes that already exist
• Recommended resources include: VRML WebSite - http://www.vrml.org, The VRML Repository http://vrml.sdsc.edu and the VRML 2.0 SourceBook
• Dave Nadeau Also addressed a request to compare VRML to Java3D. His response was that if you want to do algorithm-cs based graphics, Java3D would be the right choice. However, if you want to do quick content creation then, VRML would be what you want to use... Additionaly he pointed out that Java3D can load VRML2.0 files and give you better control over interaction than VRML can

Java3D:

I also attended both the Java3D course and went to the Java3D SIG. Go here for my Unedited Notes

• They Made a real point in the course of pointing out that Java3D was designed with Multiprocessor and Multipipe systems in mind.
• In Particular In the Java3D idea of a scene graph there is no garaunteed traversal order, and each path from the root to any node can be rendered completly independant of the others...
• Rendering is automatic, and controls the processors to make sure that there are no conflicts.
• It seemed like they spent a lot of time thinking about how to optimize the scene graph automatically... they handle minimizing all the state changes, all the culling, etc...
• Java3D ties in lots of interesting features including 3D sound and some simple animation utilities... i.e. interpolators , etc...
• There is currently a beta release of Java3D available on the javasoft web site... however it requires that JDK1.2 be installed on you machine in order to use it. Currently there is no port of JDK1.2 for IRIX systems and therefore no port of Java3D to IRIX... Java3D currently runs only on Solaris, and Windows NT/9*
• The class files for Java3D are available from http://www.sun.com/desktop/java3d
• They say there is lots of example code in the distribution that should help out in getting started...
• They mentioned the NCSA Java3D FAQ as a good resource

Performer:

Along with a lot of other people, I attended the Performer SIG. I took lots of notes including the almost complete question and answer session, so if your interested see my Unedited Notes Page. This is just a distillation of the most important stuff.

• There will be a Performer3.0 and it will be built around Farenheit Scene Graph (hereafter referred to as FSG)
• Performer will not be ported to NT... FSG is the Unit - NT cross platform framework
• Sharon Clay on SGI's implementation of FSG:
• Cross platform IRIX, HP_UX, Windows applications
• supports OpenGL and Direct3D
• FSG is the API layer and encourages vendors to specialize
• FSG does NOT implement all feature... that up to vendors
• Supports Multi-processor traversal for collisions, culling, sorting and rendering
• FSG Beta Release Winter 1998-99
• Full FSG Release Summer 1999
• There was a good Question and Answer Session given by Dave Orton from SGI see the link for the full text
• Info on Performer Releases:
• Performer2.2 is available for IRIX 6.2, 6.3, and 6.4 systems
• Performer2.2.1 is available for IRIX 6.5
• There is a version Performer2.2.2 for IRIX6.5.1m
• The Performer2.2.3 is due later this year
• 2.2.3 is promised to be a full release, while the last few releases have been more like upgrades
• SGI Announcements:
• 128CPU Onyx2 in Late 1998
• 16Pipe IR Late 1998
• New IR2 Graphics is due in August... this is supposed to be 25%-30% faster than the current IR technology
• DV Multiplexer in Late 1998
• D-Plex Support for Performer 2.2.3 a.k.a. Hyperpipe... (see the notes for more details)
• They also showed a really cool demo showing off the usefulness of the double precision transforms now available in Performer... The demo was to zoom in seamlessly from space down to a real-sized model of a car parked in the lot of Epcot Center in Orlando, FL... pretty impressive!

Show Floor

• Advanced Rendering Technology: RenderDrive
• This was a rack mounted box that did hardware raytracing using the Renderman spec from pixar... There were some VERY impressive images that they showed including effects like radiosity, soft-shadows and depth-of-field blurring... They compared the images to the same ones rendering in 3DSmax, Maya, Bryce3D, etc... the times were impressive... images computed with standard software systems were on the order of 30-40hr's while the RenderDrive image took 12 mins.
• Ascension Technology's wireless (almost) tracking
• This was Ascension's entry to the motion-capture wireless tracking market... it seemed to work ok... but it didn't seem very wireless to me...The people they had showing this stuff had sensors stuck all over there body, and all connected to what I assume was some kind of a transmitter backpack... Seemed like a real hassel to use for CAVE stuff :)
• ITU Research: TouchCube
• The TouchCube is an input device that uses the little touch pads from notebook computers... They put one of these things on each side of a cube and allow you to rotate object by pushing and dragging your fingers across the pads... Seemed like an interesting idea for a 3D input device... pretty similar to a spaceball or magellan
• PixelFusion:
• This is a graphics card from a company in the UK... They say that it is loosely based on the PixelFlow system from NC and HP... Some of the cool things that they claim are Completly programmable pixel shading, as well as True Phong shading and Bump Mapping all in real time... They said that the card will be targeted at the high-end pc market and is due to be released in mid-1999
• Autometric Inc: EDGE product family
• This is a set of tools that address the same type of issues that BattleView addresses... They are actually very cool!... They deal in extremly large terrain and mission database (on the order of Terabytes) as well as Imagery... there software and seamlessly map satellite and mapping imagery to multiresolution adaptive terrain... as well as suppporting Mission simulation modes... The person I talked to seemed very knowledgable about this type of software and market, and Actually knew a lot of the ARL people as well as the boeing people Joe Adamo, Bruce Dyke, etc... If your interested stop by my office and take a look at their excellent handouts
• FakeSpace: Passive Stereo ImmersiveWorkbench
• I also went and saw Fakespace's booth where they were demoing a new version of the ImmersiveWorkbench... This was VERY nice... Featuring a very bright display, and passive stereo... The brightness is due to using two projectors for an immersa-desk sized display... they both fit nicely inside the case... The passive stereo looked good and used circular polarization so that when you turned your head sideways you didn't loose the stereo effect..