From: broehl@sunee.waterloo.edu (Bernie Roehl) Subject: Re: TECH: My standard is better than your standard. Date: Wed, 29 Jul 1992 14:04:21 GMT Message-ID: Organization: University of Waterloo In article <1992Jul29.065416.26495@u.washington.edu> s047@sand.sics.bu.oz.au (Jeremy Lee) writes: >This is the problem that graphics people have been working on for the >better part of 20 years. There is no simple solution. I agree; if it were simple, we all would've solved it weeks ago! :-) >No, It's not really an N squared problem. To begin with, each object >only has to deal with N other objects. Second, you don't have to check >all the collisions all the time, only when an object moves. There are >tricks and tips and neat things to do that will make the problem of >colision detection less hard. Agreed. But if there are 10,000 objects in the universe, I don't want every small workstation to have to keep track of the motion of each and every one of those 10,000 on a frame-by-frame basis; even an order N problem will kill your performance, if N is large enough. We need to make intelligent decisions about which of the N objects a given object has to check; to me, this means some way of subdividing space. >And don't be restricted by a visual model. I may not be able to see it, >but can I hear it? Am I touching it? Sound is an important issue, and one we haven't dealt with (lately). As for touch... I would argue that if it's too far away to see, it's too far away to touch. And if it's completely obscured behind a wall, I can't touch it any more than I can see it. Therefore "objects I can touch" is a subset of "objects I can see". (Note, though, that I'm counting objects that are outside of my field of view but otherwise *potentially* visible as "objects I can see"). >And the universe is bigger even than any of OUR puny imaginings. I feel >that 128 bit numbers are correct. Only for worlds that need them. >I don't. Imagine the extra work you have to do converting it between >formats, even to do simple arithmetic! The conversion would be done once, at the network interface. You do arithmetic however it's most efficient *on your machine*. The overhead of variable- precision numbers is negligible; since it happens only for incoming or outgoing packets, and the network will almost always be much slower than the processor, you'll never even notice it. What this means is that for the 99.999% of the worlds people will want to build, you can have small and efficient encodings of the data. For the remaining 0.001% who really need to do quantum-level models of the universe, you provide a way of getting up to 128 bit accuracy. >Decentralization. Vote for anarchy!!! I'm inclined to agree. (Lately Dave Stampe and I have been brainstorming some interesting ideas; we'll post something when they become a little more solid.) Bernie Roehl, University of Waterloo Electrical Engineering Dept Mail: broehl@sunee.waterloo.edu OR broehl@sunee.UWaterloo.ca BangPath: uunet!watmath!sunee!broehl Voice: (519) 885-1211 x 2607 [work]