Re: Thoughts on 'the long view'
[this got rather long and full of _old_ rants, so if you don't want to see me repeat old stuff just skip to the end please] Dennis McKenzie wrote: The low poly models don't really bother me at all. They seem a part of what VR is. Ok people I needed a way in to this discussion and thanks (hmm) to dennis I got one. First of all, I will probably mention a lot of stuff allready taken up in all the very interesting posts so far, but as dennis also remarked, "Please let me know what you think" would give me my view even if I happen to share many of yours (and some not:) "Re: Thoughts on 'the long view'" These are my thoughts, = Wordy language Even though I do exactly 100% of all VRML by hand (or otherwise with Perl scripts) I wouldn't say that I'm bothered with this. I don't think this is a cause that would scare beginners off, on the contrary I belive it helps them. It helps me too. In my work I more or less have to know the Spec in my head to be able to work effecient. Yes, I use the spec from time to time, but that is only when things get very tricky and very detailed. The problem still is that it doesn't matter how I intrepeter the spec, it's how the one who did the browser and his intrepetation that counts. VRML is and hopefully will be "wordy" in the future too. = 3D is hard (NO it isn't.) No matter what medium you use it's always simple if you have the talant for it. I'm in soul a programmer and not a real 3d-content guy (even though I work as it professionally these days). VRML is probably one of the simpliest fileformats I've seen so far, with very litle specialcases and stuff in it. You don't have to know much about it to create good content. DHTML for intsance gives me much more headache than VRML will ever do since the connection and browser dependency there makes VRMLs problems like a walk in the park. Anyway, "3D is hard" isn't about tech or implementaional issues I guess (based on who said it). It's more about the concept of the media 3D if I intrepeter it the right way. This is why I was glad that Dennis gave me the way in on this thread becuse this is what I think makes the difference (but it gets tech I promise:). Forever ago I sent a mail to the list on a reply to someone that said that "all VRML sux becuase you can't do detailed stuff with it...". I got pretty upset from that mail, because he more or less told everybody that all the worlds out there where shit becuase they wasn't high detailed. Well, my worlds aren't high detailed because I never had the h/w for it. Now when I have the h/w (Octanes, etc (you mention it)) I still can't use that becuase of the viewers h/w of course. This is the point (if ever) when 3D gets hard to do. This is the point when 3D (and 2D for what it matters) gets more or less magic. First lesson for me was that I should never use textures (at least not until now). The other lesson learned was that if I don't use textures I can't be detailed. This is all very true to me since I have to remeber that everyday of work and it affects me very much. The magic for me is to create something that communicates a message and still is viewible for the users. Normaly I belive 3D creators start from something very complex in 3D studio (just an example) and than strips information (because that's what geometry is right) until it contains just enough to communicate the message. Well, I always do it the other way around. I start with one poly and adds until someone else than me can say that "oh, that is what it's suppose to be!". Then I just stop. No matter what (even if that means open holes in the geometry!). The point. I don't think that there will ever (at least not the next couple of years) be neither a format or browser that can compete with me and make this way of doing it unnessecary. But still, 3D isn't hard, not if you have talant for it (I'm not saying I have it though!). = Collision What's the problem? Use JSAI, ECMAScript, EAI whatever to do it. Oh, you can't becuase it's not stable enough or the browser doesn't support it? Well, that's I think the only problem I see with it. I did collision detection allready in Java and that worked pretty good if I may say so... = Boolean objects (ISO surfaces?) A couple of months I tried this with VRML (anyone remeber seeing a post requesting algo's for it?). I wouldn't say it's not doible, but it's damn hard. I have an implementation in C that I will sooner (but probably later) convert to JSAI whenever that doesn't trigger a securitymessage in CP. However this is to slow for almost any 3D software to do in realtime. And I tried to do a metaballs modeling feature in VRML wich is even worse. Anyway, I don't think this is really something I would need for just about anything, but I would guess the sci-guys would really like it :) = Sounds Have anyone of you tried the new java media framework? Well, we played around with it the other day and I must say th
Re: Thoughts on 'the long view'
Please let me know what you think. Remember you said that Alan ;) In a lot of respects I agree with you. I hate that I work so hard at a product that falls so short and is viewed by so few (and who knows what it will look like if they do see it), and has such a limited lifespan. VRML is dead. Sometimes I wonder if I am delusional about what I am doing. It is not VRML's fault. Oh, VRML sucks. No object collision, no a million other things that I think should be hard wired. No STEREO vision! No GLOVE support! This is VR for God's sake! Where's the hardware?? But it is not VRML's fault. It is the content creators fault. It is my fault. I have to find a way to make my VE's *Immersive and Interactive* without neural implants or holodecks. It's slow and not as pretty as a game. (I'm not spitting John :) Well it could be as pretty as a game, but then it would be waaay slow. Guess this means we are going to have a limited audience. Unless we could create something very compelling. And I have very few clues how to go about that. More clues please. Software. Kinda beta software. Not that many are willing to put up with it. I'm not that fond of it myself - and I'm addicted. I'm not that sure that there would be that many more VR enthusiasts even if the browsers were perfect. They gotta have someplace interesting to go. Maybe the real attraction to VR is the act of creation itself, and the only people who are going to be interested in what you do are other creators, and then only as it affects or compares to their own work. I hope not -- this is kinda my worst fear about what I am doing. The low poly models don't really bother me at all. They seem a part of what VR is. Imaginative worlds are key, not high realism. I've seen some low poly worlds that turned my crank and some high realism ones that turned me off. I thought Bit (the star thingy in Tron) was a pretty good character and his polycount wasn't much larger than his vocabulary. That would be cool if we could make movie quality models for our worlds and get good fps, but we can't and I just don't feel that it is that important to VR acceptance. Four polys or four thousand our characters and objects need to be given realness by context and actions, not necessarily by glitz. Very, very, little compelling content (actually none that would make a person not interested in VR take up the banner I think). A fair amount of stuff that makes me think that if the hardware and software were in place -wow! where could this concept be taken. Lots of things that wouldn't be noticed by someone outside of the community or of our artistic bent, but are cool to someone who is looking down the road. Building blocks. I gave up the vrml-list because I'm darn near computer illiterate, basically learning only as much as I need to know to satisfy my habit. That's why I'd love to see this list more active (you all doing all the work in that regard of course :) I hate the tech end of it. Robert, is there any way to wire VRealm directly to my imagination? (How embarrassing if he can do it and I still fall short!) I don't want to step on any toes, but I was never much of a fan of SGI. Being a realist - thanks though SGI for taking it as far as you did. And I sure miss buzz.wrl. (Even though half of my submissions got rejected by judges who obviously didn't know a great thing when they saw it. What was the deal there Jed?) ;) Buzz was a _good_ thing. We all gotta have some strokes now and then. Appreciation and a high paying job are nice strokes. VRML has little to offer by way of these though. To Alan from a worldbuilder: I question the viability of this medium and my ability to work in it all of the time. Sometimes more harshly than others. But in my heart of hearts I believe there is something special about it. Having this feeling gives me the ability to tell when someone else has experienced the thrill of 'playing God' by looking at their work and the care they take in it. It is pretty obvious in you. It is depressing to see perhaps the most talented member of our fraternity write a letter like you did, but it is understandable. I'm pretty sure we'll be seeing you down the road though. Its doubtful that anyone could give it up forever after taking the deep drink you have. I sure hope not. Dennis
Re: Thoughts on 'the long view'
Bob wrote: Just start off with the idea in mind that you don't know what events are -- that they *aren't* like anything you're familiar with, and you'll get it. And then you'll get really annoyed with the people who designed it. That event-cascade thing has some incredibly annoying side effects -- when I figured out that I needed dummy ultra-short-duration time sensors to generate a new timestamp to ensure a specific ordering of events, I was appalled. And that stuff was finalized after the Handbook went to press, so we don't even discuss it. Sigh. As for realtime raytracing, there were rumors at SGI back around the start of the VRML 2.0 effort (so call it late '95) that the next generation of SGI's rendering would be a realtime raytracer. Never went anywhere, and the hardware to do it still isn't cheap enough for public consumption. But graphics folks have at least been considering it for a while. --jed, doing a lot of wandering down VRML-memory lane this week
RE: Thoughts on 'the long view'
Miriam wrote:
* I would like to see a simpler, less wordy language.
I totally agree with Cindy: the problem here is not the wordiness of the
language, it's the dearth of good authoring tools. Nobody should ever have
to write {VRML, X3D} by hand. Top creators will probably still have to
tweak it by hand, but wordiness-of-syntax is much less of an issue for that.
quoting John:
Certainly "appearance Appearance { material Material { ... } }" is ugly as
heck and should be immediately killed
I remember the meeting in which that syntax was chosen. It was Gavin,
various other VRML designers, and the Performer engineers (and me and maybe
Josie taking notes for the Handbook and the spec). The Performer engineers
convinced the Inventor engineers that separating things out that way made
certain important optimizations possible in the browser. Everyone was put
off by the syntax, but the winning argument was that authoring tools would
soon hide the syntax. Unfortunately the authoring tools took a long time
to come to fruition, and still aren't as far along as they ought to be.
But that's the direction we should be heading.
Think about it this way: imagine a world in which images have to be
edited by hand in a binary editor. You have to fiddle with binary numbers
representing strings of pixels of different colors. You might say "the
binary representation of images sucks!" But the real answer is PhotoShop.
* Allied to the above point is routing and interpolation.
Repeat after me: Authoring tools. Authoring tools, authoring tools,
authoring tools. It's a good mantra.
(Cosmo Worlds uses a "Keyframe Animator" to take some of the sting out
of interpolation. It's not all the way there yet -- it has rough edges and
non-intuitive aspects. But it's way better for most people's purposes than
typing interpolator numbers by hand.)
(I do, btw, like your idea of vectors as properties of objects. The
VRML event model has a *lot* of rough edges.)
* a general solution to the gravity problem
I played a Mac version of the arcade game Gravitar once; you could
design your own levels, which involved placing gravity generators in
various places. I rather liked that approach -- very simple and intuitive
from the content creator's standpoint. "This point is a source of gravity;
things will tend to fall toward it." For simplicity of VRML worlds smaller
than Earth, we could have DirectionalGravity instead of PointGravity...
* a general collision solution
Whether or not the physics engine is in the core spec, for our purposes
we *will* need a physics engine. I saw quite a nice one at SIGGRAPH '97,
but don't know how much it succumbs to the problems mentioned on the main
list (about chaotic systems and such).
* Boolean objects. From a content-maker's viewpoint they would make life a
LOT easier, but I think it might slow the player software a lot. Perhaps
that has to be relegated to just the sculpting software.
Yah. There's a name for this, but I forget what it is. We'd love to
have it be part of Cosmo Worlds (it's one of the things that designers
always ask for), but it's definitely not gonna be in 2.5.
I remember a post about a year ago by somebody who managed to make sound do
the doppler thing with a subway train approaching, passing and going
away..
I just saw this again the other day. I think it's at
http://www.janvier.com/ -- but that may just have been slowing down the
sound as the train arrived.
So how do we want to present our list of demands^H^H^H^H^H^H^Hdesires?
Should we compile a big wishlist and just forward it to the Consortium? Or
should we try for something more formal? Perhaps get the CDWG folks in on
this? (I dropped that list for lack of time some months back, but it does
seem like the sort of thing they should be discussing too.)
--jed
Jed Hartman*[EMAIL PROTECTED]
Wordplay column: www.kith.org/logos/words/
Re: Thoughts on 'the long view'
Wow - great discussions going on here, cool. One thing I wanted to clear up was something that I guess I didn't get across too clearly - I'm not a non-technical guy, I'm a non-bureaucratic guy - Heck I make my living programming. When I said I get lost in the technical jargon on the www-vrml list, that tends mean the glassing over of my eyes when I see a post on the theoretical application of XML to a 3D namespace, or an argument on the scope and style of a spec for some acronym I'm unfamiliar with, etc. The VRML spec isn't what I'd consider too hard, especially not to create something fairly basic. Where it falls apart for me is at the edges - when you want extra detail, or some level of realistic rendering, or a high level of interactivity or logic. I can see authoring tools clearing up some of that, but as Jed said, I would always expect some hand-tweaking to be necessary to get _exactly_ what I'm after. I agree with Cindy - 3D is hard, but once you can cross that barrier, I don't think the spec is that much harder - it takes a while to grasp, but hey, what new language doesn't. I find myself thinking in X,Y,Z unconsciously now - often, imagining how I might represent something I see in the real world in a 3D program or VRML world. I was floored back when I found out that the only realistic way that developers would have access to the scene graph from an external source was through Java. Java had some promise then, but I hated it then and I hate it now, and the promise of it isn't getting much better (IMO). Applets are nearly as untrustworthy as VRML plugins and the EAI has always appeared to be an ongoing experiment-in-progress, an alpha release at best. I also didn't want my rant to appear so anti-VRML as it was. I have a lot of problems with VRML - but that doesn't mean I'm not still hooked as a developer, I just hit a wall that made me decide to scale back and play the field a bit when it comes to delivery media. While we're discussing new formats and the future, let me share some recent experience I had with Microsoft's Chrome(effects) Being on the Microsoft Campus, I had an opportunity to actually use and create some Chrome content, using the XML syntax they had to create 3D objects, etc. I don't know how many of you had a chance to play with it, but ugh... let me tell you. If you think VRML was bad - here's something that has all of VRML's problems - worsened, plus a few more. The only 2 things radically different is that it is a)called out by XML in the HTML page, as opposed to a separate file in an embedded plugin and b) can render real HTML as an Imagemap. It is no wonder to me that they have shelved it for a while. 1. It's slow. Did I say slow, I mean glacially slow. I don't recall directly measuring fps on anything, but can tell you that I never saw faster than about 10fps - usually closer to 2-4 fps on a 200mHz machine with no acceleration. I once saw some work on a prototype 400mHz machine with video acceleration - it was still so jerky and slow I was truly amazed. I've been told it's because of the many layers it goes through (from page to ActiveX component to DirectX to base hardware, and back up the the to the display), and the general slowness of DirectX. 2. It's limited to a few basic shapes and using .X files for custom objects. The only way I know of to create .X objects is to use 3DSMax and a plugin. How many of you have $3000 to $5000 to spend on a 3D authoring tool (let alone the hardware to run it)? I tried many many translators, and had very limited success, with no image mapping info ever being passed on thru the translation. 3. It's (projected) general audience was so limited as to be laughable. Only those using a machine from a major manufacturer (like Dell) _AND_ running Windows98/Win2000 _AND_ a 400mHz+ Intel processor _AND_ Microsoft Internet Explorer 4.0+ _AND_ with a very certain small set of video devices drivers installed _AND_ with the ChromeEffects component installed could actually access any of this. This automatically left out everyone using a Mac, using some form of Unix, using Win 3.x, using older machines, those who hand-built machines, those who like a non-compatible video display, and those who bought machine from a non-major manufacturer. Such a small audience for such slow performance. Working with XML to create the 3D content was weird, yes, but not necessarily any harder than VRML was at first - there's a lot of adjustment to do when it comes to units and distances, and Chrome felt a lot closer to 2D sometimes than 3D. I always looked at it as "VRML's 3rd dimension takes you 'into' the page, Chrome's 3rd dimension takes you 'above' the page". Whatever the major drawbacks were to Chrome's first try, I don't think it's related to the XML syntax, quite the contrary, XML as a method of description is very cool, since it makes scripting 3D from the page much easier (without needing the godawful EAI
Re: Thoughts on 'the long view'
My thoughts on the long view are this. 3D is the future of web design. It will come to pass. I will to do my best to towards the effort to make that happen. If this means working within the limitations of the vrml plugins, and existing hardware, then I will do that. The future of 3D is dependent upon those of us who are willing to work within those limitations to push the envelope of content. I believe our imaginations, and thus our demands on this technology should always exceed the scope of any new vrml standard and the capabilities of even the best vrml plugin. Richard Kapuaala
RE: Thoughts on 'the long view'
BTW: per Gavin's input about SVG and long view thoughts. Just for the heck of it, go to http://www.w3.org/TR/WD-SVG/struct.html and see how many concepts you recognize instantly from your work with VRML. This may go a lot quicker than anyone thinks. Len
RE: Thoughts on 'the long view'
It may or it may not. The big impact in the beginning will be a different model for the style (eg, if they use CSS as they do in SVG). For story telling the situation may be better. We've spent a lot of time trying to figure out how we should do branching, interactivity, etc. here for these stories. One problem has been the model where the world was an embedded thingie inside a web page that had to communicate with the browser which acted as an intermediary between the world and the operating system then the web. It could be that if the 3D, 2D and text objects are treated equally on the browser page, things get better. What if we get some things we need from the fallout o Standard event models for interactivity/behavior. Right off the bat things are better for onmousedown, onmouseover, etc. The script model looks just like HTML, VisualZed, etc. o Same reuse of scripts as in HTML pages o Reusable styles o EASY integration of text o Same support for streaming audio as page has and possibly the same synchonization model. IOW, it is possible (keep fingers crossed) that we could get a quick improvement in the tools and in the medium to long term we get REAL broadcast formats. Looking at the very quick movement in broadcast devices (eg, the flat panel displays, the increasing convergence of satellite and cable), knowing people are starting to look at the idea of just buying a big PC monitor, a DELL and putting that where the living room TV is. In that environment, for our interactive stories to work, we need broadcast worthy protocols and formats NOW so we can learn to compose for them. We have been writing stories in terms of the world model. Maybe an integrated page model is a step backward, or maybe it is an easier model to work with. We'll see. Whatever we have learned from VRML97 should be useful just as the things we learned from 1.0 helped us in 97. Len Ekam sat.h, Vipraah bahudhaa vadanti. Daamyata. Datta. Dayadhvam.h -Original Message- From: Jed Hartman [SMTP:[EMAIL PROTECTED]] 7. Unfortunately, points 5 and 6 are likely to combine to result (in the short term) in VRML being optimized as a medium for ad banners and visualization tools.
Re: Thoughts on 'the long view'
Alan wrote: I don't exactly know the technical reasons why VRML is so much slower to render realtime than proprietary formats in games, but you know what - it really doesn't matter. This has been a very difficult issue for me to accept. I know some of the reasons VRML renders slowly, and my approach has thus far been to say "it therefore behooves content creators to create content that works around the limits of the technology." Which is good as far as it goes. But in the end Alan is right, our audiences don't care about any of that -- they just see that VRML isn't as fast or as slick as Doom, as complex or as shadowy as Toy Story, or as pretty as Myst, and they say "Ho hum, why should I care?" The problem is that the reasons VRML renders slowly (particularly its generality) are required for the kinds of things we want to do. Sure, if we accept the Doom model we can tell interactive stories in beautifully rendered fast 3D spaces -- but only if those spaces all have a ceiling and a floor, and rectilinear walls, and conform to a set of predefined behaviors. Say you're someone like Picasso. You go down to the art gallery and you say "I'd like people to buy my paintings." The gallery people say, "Okay, paint me fifteen pictures of boats by tomorrow and I'll sell 'em all." You say, "But wait! I don't want to paint pictures of boats! I want to paint pictures of nudes descending a staircase! Oh, and I can't paint fifteen of 'em by tomorrow; my art requires time." The gallery people say, "Sorry, there's no market for your work, better luck elsewhere." This clearly isn't an exact analogy. But it does show that some of the problems we're facing here aren't unique. Everyone could tell Picasso "you're just making excuses; the fact is that your approach to art is slow and doesn't live up to the beautiful boat scenes that other artists can turn out." And from a commercial perspective -- or even a popular-acceptance perspective -- they're right. The main difference between that situation and ours is that we agree that what the public is getting *is* higher-quality (display) than what we can give them. But that doesn't necessarily mean it's better art. It seems to me that speed is largely a technology issue. When we have faster rendering, we'll be able to do VRML at better speeds. We'll never catch up to the game market (this is where I disagree with John DeCuir's attempts to make VRML a viable twitch-game platform), but once we can display complex worlds at 30fps, we'll be able to pursue artistic and storytelling goals without worrying about whether our special effects are as good as Doom's. (Look at early black-and-white movies -- they certainly didn't look much like reality (no color! jerky! no sound!), but people liked 'em anyway.) The other approach, of course, is to take what works and try to adapt it to our needs. We can try to use the Doom engine to tell a compelling story. We can render a series of still frames from 3DSM and produce Myst-like games. We can do an awful *lot* of pre-rendering (and use impossibly large amounts of disk space) and try to create interactive movies that look like _Toy Story_. All of those things are viable options while we wait for VRML to get where we want it to be. There are other issues related to public acceptance as well. F'rinstance, I just finally got a friend of mine who runs the Information Science hall at a major science museum to look at IrishSpace. Her first comment was, "I liked it, but it's not very interactive." (I had warned her about that months ago, quoting the "radio play with VRML illustrations" line, but I guess she forgot.) She wondered what the advantage was of doing such a story as VRML rather than as a pre-rendered 3D movie. I think that the interactivity that is there adds a lot, but I don't know if I could convince my friend of that. 2. "It runs unpredictably" Is this any more of an issue for VRML than for Java and JavaScript? There are a lot of platforms out there, with a lot of unpredictable combinations of hardware and software. Even a company that does extensive cross-platform testing can't hope to hit much more than the most popular combinations. This will get somewhat better when/if one browser interface becomes better distributed. If Cosmo Player replaces WorldView and can get broader distribution, if we ever ship our Mac version, if Chris Fouts finishes his IRIX port, we'll be able to count on things looking and working much the same on a variety of platforms. But that's a lot of ifs, and relies heavily on one company (while ignoring other good browser options out there), and still doesn't cover things like Linux and older Windows systems. And X3D may mean starting the whole ballgame over again in terms of browser distribution. The logfiles I've seen tell me that anywhere from 1/2 to 1/3 of the users who hit our ISS intro page go in and actually get the model file (one more click
