why only primitives?

Does AW use Havok? Bounding boxes do not make a very compelling interaction. It really harms the immersion and confuses users.

My few experiences with AW is that it's a very poor environment to interact in with terrible graphics, however I may not have gotten to the right places in World.

course twist/hollow ect add more. I was trying to be very conservative about the polygons per prim as to be fair

He probably means only for the initial pass of the physics engine, once they collide it would be done per-polygon.(like it is now)

ps.. perhaps people should state there knowledge in the field when they reply so we know how to answer them? -- Personally i've made own OpenGL 3d engine, but it was only as a hobby, so I wouldn't call myself an expert. but I know a thing ot two

How's this for mindless repitition, you are the but one of a hundred people that have suggested this since beta TWO YEARS AGO. Is it here yet? Hmm looking around I don't see them putting priority into into and you keep bringing active worlds into which is a completely different system with relitively little physical interaction(there is wind and other external forces in SL that the sim has to keep up with). Now here are some numbers of my own, for 1024x1024, high quailty JPG files, they average around 500K now when they hit server side they are converted into JPG2000 format so I'm not terribly sure on the final file size. But for shits and giggles I'll use half that so 250K per texture, now 2000 textures textures per sim since having a standard texture set is a rare thing lately, and you get roughly 500 megs of data right there. Oh my where did that come from? hmm now for the model part, I chose the OBJ format for shits and giggles, it's an all text format file type that references out for materials at least from maya, 3,000 poly object is roughly 500K as well. Now can your bandwidth and more importantly their bandwidth handle roughly a meg of data output to a single user per object? Once agian I may be making great assumptions with model size but I thought I'd show the extreames as in a potently abusive situation, there are a lot of texture artists out there that still use 1024x1024 for "quality" so you've already got a huge bottle neck there, let alone if people start adding "quality"with thier models. Your points mean jack, this has been discussed more times then can be imagined and shot down every time for two years, your ideas aren't going to change solid facts till the engine makes some changes.

Personally I didn't like AW...I felt it really lacked substance, models though poly seemed too simple, all the ground I saw was flat(SL is generate from .RAW files) and animation if any seemed terribly crude, but then agian I may not have seen all of it.

PS- So prim modelers should price thier work lower because they can't understand/afford a 3d package? Even if they put the same about of work in, sometimes more for those of them that model using trig equationes rather then "well this looks good".

PRIMS aren't our problem in SL. It's the interaction with them.

I'll be happy for the NEXT 5 years dealing only in prims as long as LL fixes a dozen other aspects of the whole universe. Why so agry about prims? I'm angry about the cheap-ass editing software. I realize anyone who has modeled before will be spoiled but at least give me an editor that recognizes edges, corners and middles.

I realize this is asking the engine ot treat my prim as a series of lines and nodes, but honestly it's getting very rediculous as far as alignment and placement is concerned. If i could somehow convince LL to give the good people at @Last Software a call, we'd all be enjoying a small, newbie-friendly editor like Sketchup for handling our prims.

Not being able to snap at the edge of another prim or grab that prim's edge at least in the middle is killing the production value of builds. Beutiful builds showing no comprimise are KILLED because of gaps and misalignments. I live in a house made simply of cylenders and domes and it's damn near stunning ONLY because i chose a setup that had absolutely no right-angles to fail and no two prims that needed to apprear as one large one.

~Lefty.

I did'nt like Kensuke numbers as they had nothing to do with the inner workings, so I worked it out.. I was gunna post the whole calculation but its so big, lol.. So what I found using a cross in SL is it would take 2 prims and would take about 240 bytes.. and to reproduce it in polygons took 432 bytes without a UV map. So in short Prims are more efficent then I thought. But that was also a case favorable to prim modeling. Id say size wise poly modeling would take up more. The benifit though of course is more complicated objects and faster rendering speed client side.

Something I still think that would be sweet is nurbs based surfaces. They would be a squair thin surface, with probably a fixed number of control points. when you move each control point it would push or pull a curve into the surface. if they could be lined up you could build whole objects out of them. question with these is how fase can they be calculated in real-time..(actually.. it would only need to be calculated one it came info view)

hmmm.. stuff

Actually im pretty happy how they have done it. Id say quite well. Something you need to do if you dont want gaps is to learn to build by numbers. That just means cut and paste x values, ect.. You also can use the snap to grid to deal with this.. which it great for making houses and that sort of thing. Also SL uses a LOD system based on distance. If it looks fine up close and see a gap far off then theres not much you can do. Its a limitation of SL. (I wish they could maintain edges at all LOD levels)

I enjoy the scary numbers myself >_> but thats me. Prims at low levels are more effective, once you get into higher situations it might not be the case rendering wise but as far as data sent to the user it would be, it's the balance between the two lag points, either you get lag from downloading or you get lag from rendering it all, lag rendering side is the more bandwidth friendly choice for a company still building itself.
As for the editing system, I do think they could add things to the prims that would make them better(like they did with the last prim update), if you want the editing to be on the dot though you almost have to use math, or have them up the snapping grid size.
My numbers came from a smoothed version of SL avatar chest that was rigged in maya(just the chest no arms)and the texture size was the one wrapped over it btw, the accual files for the SL upperbody are about 1 meg in total size, but that includes the many different morph targets it uses.
PS-Without arms the SL av chest is 760 polygons(1362 triangles), 96.5 KB in the OBJ file format with UV map

Always good to hear from an expert.

Just a quick question - how does havok handle bounding box heuristics when it comes to complex polygons of this sort? Maybe we could get some of the local experts to create some how-tos here and the lindens could read it and get some ideas on how to enable this behaviour.

I am actually outright against importing meshed items. Now, I can use a modeller so this has no effect on me, but the ability to import meshes with drastically increase the quality of most things in second life. Now this sounds all well an good, right? Well, it is my experience that the more complex you allow something to be, the more peopel bring it to the limit and the few people actually attempt to make anything.

As standards increase people stop trying, save for the elite. If you want to use a modeller I'd recommend getting the There development package.

Atop that -- though this is a personal thing... I like to work within limitations. It's both challenging and makes it more accessible. Besides, you can do some AWESOME work with Primitives. You just have to learn the tricks.

I'm not sure if you mean this sarcastically or not, lol.. Ill assume not.
Also I'm not an expert, especially when it comes to havok. I don't know exactly how havok works, I wanted to download some doc's on it a few weeks ago but you have to sign a NDA agreement and its a big hassle.. So I can only assume how it would works. Actually something I'm not sure of is if it checks collisions on the poly level of the meshes or if it does it in terms of mathematical shapes. (that would be tough on the spiral, lol)

in any case the existing box prim could be used as a bounding box for a poly model. or maybe havok(or havok 2) can handle it natively... I'm sure it does, but if collisions are done as mathematically can they mingle. I'm dono.. Anyone actually use Havok here? lol..

AW doesn't have any physics engine other than avatars (as I described in another reply in this thread). However, AW's collision detection is pretty detailed (though slow).



Depends which world you go to. AlphaWorld, the largest world, is crap, but AWGate looks pretty good (though SL looks better).

Kensuke, again you're failing to grasp the concept of EFFICIENT polygonal modelling. Any idiot can create a bajillion-poly object with obscenely-sized textures, but I'm not talking about that in the slightest. Stop focusing on what YOU think I mean and actually attempt to understand what I am actually writing, because I tire of dealing with your broken recordness.



Depends on the world. I've seen quite complex (1000s of polys) objects (not including most avs which are already 1000s of polys) in AW, but builders soon realize not to abuse polygons.



AW's ground depends on the world but most have non-flat ground now. In 1997 I made hilly modular ground in Cubed world, so it's been possible since AW's inception (1995). http://tnlc.com/rw/rwx.html#ground for more info.



<shrug> Prim modellers can price their work however they want, but if someone chooses a more efficient, optimized, poly model over an inefficient, unoptimized prim model, you do the math.

If Linden Labs is worried about bandwidth they should allow modellers to host objects, textures, sounds, animations, etc on their own servers, like Active Worlds does. Why should LL have to host everything? Hell, allow sims to be hosted on other people's servers too, like AW's worlds are.

I understand the concept of effeciant polygon modeling, in fact I assume that modelers do, but from a business perspect you have to look at it that the average user does not, you have to look at the other side of the aurgument before jumping into things here, not all users are going to magicly decide to play by a set of rules because you think they will. By the way I did find one of these non-flat worlds in AW, took a bit but I did, I fell through the ground that for some reason had not yet loaded and noticed one thing...AW doesn't appear to have a lighting or shadow system.. O.o it does have a nice flat sky though and at least in the version I'm using have a maxium of 200 meters in the viewport. It also took ten minutes for the zone sounds to hit..

As for why LL doesn't let players host their own servers, player servers have tendence to go down, as well most would also have trouble trying to handle the bandwidth load that SL requires less the player goes out and grabs a static T1 line, but most people can't afford that so they'll try hosting a sim on a DSL line or a Cable connection, which would be all fine and dandy unless it went down or the IP address was changed etc. The reason is more a method of control then anything else, with all thier servers housed in once place they can be easily regulated, half the grid doesn't dissappear at random, and well they plain out make money, they are a company thats what they want to do, pretty much the same reason SOE and Blizzard don't give out their server software. This once agian is another well talked about feature, which would be cool and hell I'd be for it but I don't think it will happen anytimes soon.

Because AW is really raking in that dough... if you're going to suggest that the Lindens scrap their business plan, I highly reccomend that you suggest a profitable plan. By decentralizing servers, what does LL gain? What does LL lose? Let's make a simple table.

PROS:
Drastically reduced server costs
Reduced technical support costs
Some amazing sims being run using 128 node Beowulf clusters

CONS:
No chance of making any significant profit
Loss of control over content
A horde of horribly equipped servers which everyone will complain about (Even more than now)
Anarchy
Huge potential for IP theft and explotation

I'm not sure Eep is necessarily saying host everything, just the models.

In the end, I really think the problem isn't the bandwidth, but the interaction with the prims. However, it would be interesting to hear from a Linden about this .. maybe Uncle Linden will get us some more information.

It's too bad that LL doesn't try to leverage the collective wisdom of its userbase more often.

The real point behind this has been made, but I'll make it again.

The ppl who want poly meshes propose all these changes to the social structure, the economy, the physical mechanics of SL, just so they can have nicer 3d models.

Who cares?

That's not SL. The only thing that makes SL worth playing is the people. SL is a social engine, like a muck/mush/moo, with a 3d environment. I'm sure the graphical quality will improve, when it can do so without destroying the true underlying structure of the human interactions that makes it worth ppls time to build things in the first place.

If poly modeling were to come to SL, I suspect a lot of ppl who are integral to the world as it stands now would either have to or want to leave. I suspect I would too.

If you want a place where ppl can walk around and be impressed by each others leet modeling skills, well, there are other places like that. Why go here then?

Why do you propose limiting SL's growth as an interactive virtual environment? There are probably ten thousand other ways which humanity has devised for communication. Why use SL if it isn't for the graphics? If all you want to do is talk, then go give semaphore a chance.

SL exists as a place to communicate, to build, to code, to play, and to do anything else which we feel needs to be done. To argue that improving the graphics/models/physics of the game will ruin it is ridiculous. If you aren't here for the modelling, then don't model! You can still talk about sl00ts or whatever, regardless of whether other people are modelling.

Shame this is all a moot point anyway.

Indeed, but it is something to do during IS103 which is now over, woo!

Look in the mirror, Kensuke; I could use the same argument for your assumptions that modellers WON'T model efficiently if they have the tools to do so. I am willing to bet more modellers WILL learn to model efficiently.



Again, it depends on the WORLD (not zone). AW does indeed have multiple, dynamic, colored lighting (vertex). The only time you should fall through/before ground is loaded is if a world isn't using an infinitely repeating ground object or terrain ground.



<shrug> Then LL is shooting themselves in the foot for NOT allowing remote servers (optionally, of course).

nope, its a MOO point...

MOOOOOOOOOOOOOOOOOOOOOO!!!

I have, it's very pretty on the other side of the mirror, all flowers and sunshine(The glass is indeed half empty btw).
Ahh this was one of the worlds right of the main..hub..thing whatever it's called, I didn't see much evidence of lighting but once agian you have to look, just like in SL I'm guessing, SimHorror was an awesome example of what you can do with prims and decent texturing.
In the long run they may release server software but I'm guessing thats years down the line, right now they'd be shooting themselves in the foot if they did for reasons previously stated but anyway...

apt-get moo