From: almquist@brahms.udel.edu (Squish) Subject: VR OS wish list (was Re: Update rate, and the C-words) Date: 8 Apr 91 02:42:11 GMT Organization: *STUCK* at UofD for 2 mo' months In article <1991Apr8.000801.8080@milton.u.washington.edu> (Bruce Cohen) writes: >The state of the art in physical simulation of mechanical systems >(bridges, chains, snakes, and jello) involves solving a set of >simultaneous differential equations everytime a part moves or the >forces on it change. In my book that means floating point, and lots >of it. How does this thought sound? Why don't we create a virtual reality system based on a efficent computer solution that has been working for a while? What could that be? How about Operating Systems? Why don't we create a VR OS - ie. a kernel that would handle in addition to usual operating system responsibilities (ie. memory management, file system, etc.) it would handle VR related tasks (ie. high speed I/O information, 3D sound, interactive 3D graphics, etc). Nothing new with this idea, many people have come to the same conclusion. Taking from previous posted articles of yester-year, there was someone who said that they wanted to be able to have sound coming from ttyX and the graphic texture map info coming from ttyY etc. Using this concept why can't we get physical simulation info from ttyZ? If I/O is being established between VR kernel and DataGloves, why couldn't we establish a connection between mathematica and objects in our VR world? We could do bridge design by establishing a link to AutoCAD and then do bridge analysis by establishing a link to SAPP90 or Strudle or Mom&Pop's FEA (Finite Element Analysis) package. ETC. I've seen a bunch of neat packages out there on the market. Why not use them? Or should we follow software practices of yesterday and re-develop the wheel (-: SO, would this work? Well, if we had our simulation package running on a CRAY we'd be in busniess BUT, what about parallel and distributed processing? If we designed our kernel so that communication is fast and efficent and also so that it would allow multiple connections, what would stop us from using various concepts from the parallel and distributed processing schools of thought? Also, I've heard rumors that (was it Macsyma or Mathematics?) was going to be put on a chip!! There is also the new chains of RISC machines that are pushing the computational envelope - how about the i860? What other creative/dreaming ideas could we use? What kinda abilities should we implement or consider putting in this wonderful VR kernel? Anyone got ideas? I see a VR kernel coming to be in the future that would allow wonderful physical simulations to be a common thing. Using this idea, we wouldn't have to have everything running in/on the VR kernel. Just establish a high speed link with the kernel, talk the same language, and TA-DA, we could use UNIX machines, DOS machines (ARF!), VMS machines (YUCK!). Maybe now we might be able to get some use out of your BIG Dinosaur, aka the IBM 3090 (as taken from Bill Joy's comments about this wonderful machine)! Toss out the debate about Cray and Connection, you could use them both. Only problem, we'd need GIGA-networks SOON. But with recent breakthroughs in fiberoptics, who knows. SO, where has this posting lead us? What's in YOUR wish lists? What kinds of abilities would you think a kernel should have? What kinds of problems could we solve? How useful would our cheap VR systems be? How could we make them more useful? Inquiring minds wanna-know. (-: - Mike Almquist (almquist@brahms.udel.edu)