[sldev] [ATTN-LL] Physics engine status update?

[Nik Radford]

----- Original Message ----- 
From: "Dirk Moerenhout" <blakar at gmail.com>
To: "Argent Stonecutter" <secret.argent at gmail.com>
Cc: <sldev at lists.secondlife.com>
Sent: Thursday, September 20, 2007 6:22 PM
Subject: Re: [sldev] [ATTN-LL] Physics engine status update?


> On 9/20/07, Argent Stonecutter <secret.argent at gmail.com> wrote:
>> From: "Dirk Moerenhout" <blakar at gmail.com>
>> > Meshes are solely a way to represent the objects graphically. This is
>> > exactly why sculpted prims have dodgy physics. There's no way to
>> > represent it mathematically so the physical object is not at all alike
>> > the visible object.
>>
>> Sculpted prims have dodgy physics because the sim is not using the
>> sculpt texture to calculate their shape yet. Whether the mesh is hard
>> to simulate or not isn't relevant... the mesh isn't actually being
>> used at all by the sim. From comments Qarl has made they are
>> intending to use the actual mesh at some point, but right now all
>> sculpted prims are treated as spheres by the sim.
>
> You have rather strange logic and I guess you didn't get what I was
> implying. I know it's a sphere so let me put it this way:
> They are using a sphere to represent it because it's too complicated
> to simulate the mesh.
>
> If you truly believe the mesh complexity does not matter then I'd
> really want to know what you consider the reason to opt for a sphere.
>
> As for using the detailed mesh I wouldn't take bets on it if I were
> you. You can use it if you downsample it but that it'll remain a rough
> representation. Best bet is to calculate Bezier curves that match the
> shape best. But still the most important part is: you need a
> mathematical representation, not a triangulated result.


>
> Dirk aka Blakar Ogre
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html

Triangles are often used when needing exact collision. Maths can be used on 
two meshes of triangles to determine collisions. It can all be 
mathimatically represented. The triangles are the data, and mathimatical 
function are applied to that data. Just like maths is used to display the 
triangles, maths can be used on the triangles in any number of ways.

Simple bounding boxes and spheres are the simplist of collisions because the 
representation of them is a considerable less effort to calculate for than 
meshs. But meshs can be, and are used in collision detection, the can be 
used as to calculate the theoretical centre of mass for an object, assuming 
each part of its volume has an equal mass.

Think games like half-life, where a wall has been modeled in the world to 
have a odd shaped hole blown through it. Though the players collision is 
treated as a bounding box, the wall is not. Collision on BSP's pretty quick 
to determine due to the nature of being able to singlw out the vertexs, 
vertices and triangles involved in an extreamly quick fashion, but its still 
done on triangles.

If a physics engine wants to accuratly simulate collisions on an object it 
will do it, on triangle basis. Perhaps originally using bounding spheres and 
bounding boxes, to narrow down the triangles it should check, but it will 
still check them.

By the way, I'm not aiming this at anyone. And I apologize if I've appeared 
to ramble