[vos-d] web3d symposium report

[Karsten Otto]

Am 27.04.2006 um 14:35 schrieb Benjamin Mesing:

> I don't really know A3DL, but could you provide a rational why X3D is
> not suitable, or A3DL is the better solution?
>
As I understood VOS, it is not only intended for human users, but for  
software agents as well. In this case, the issue is related to scene  
semantics. Also, it is not limited to X3D alone, but to most scene  
description languages. Consider a simple example: For our virtual  
park, we want to design a simple tree.

In VOS, we create a new metatype park:tree, and provide an  
implementation for everything you might want to do with a tree. We  
place it in the park by giving it a a3dl:position child, and adding  
it to the sector. We also give it an a3dl:model property and  
associated a3dl:material that contains some visual representation.

In the scene language, we construct a small scene graph instead. We  
place the tree in the park by linking the parks toplevel node to a  
transform node. The latter has a vertex set child, a polygon child,  
and a material child, which together create the visual  
representation. We also may associate a mouse listener and some  
script with this data structure (likely the transform node) for  
interaction.

For the human user there is no difference between the two, because  
the user has the cognitive ability to recognize color and shape of  
the visual representation as being a tree.

For an agent, the scene graph is just that - a generic data  
structure, containing lists of values. It has not cognitive ability  
to interpret this data as being a tree. Also, it has no GUI client,  
and thus cannot interact with the tree, unless it uses some nasty  
hack to simulate mouse clicks. In contrast, the agent can recognize  
the VOS tree by looking at metatypes, and can interact with it by  
calling methods on its proxy.

Regards,
Karsten Otto