This is probably the best deffinition of collision. Which brings up another
point, some are talking about collision as a behavior of objects. How do the
objects communicate their position to each other?
Possible models:
1 Object "a" polls moving objects to see if they collided with it.
2 Object "a" makes a bounding box check of each object, followed by (1)
if a collision occurred.
3 Object "a" knows where other objects are going to move, and plans
a collision with another object.
Another issue I might raise is about collision in hirearchial models, which
may make for some messy mathematics, and long computations.
Consider:
Two knights fighting in an interactive game.
(Knight in armor) owns (knight arm) owns (knight hand) owns (sword)
colliding with:
(Knight in armor) owns (knight hsad).
(A typical occurrence in such a game)
The natural question to answer first (in my opinion), is should
collision detection be a behavior of the two nights, using their
own specialized algorithm, or should a general algorithm be
evaluated for the nights, or for their parts.
>
> My suggestion for an approach to behaviors which are common and must be
> FAST: Get the behavior very standardized and implimentable as scripts.
> At this point browser writers, to improve their browser performace,
> should internalize the behavior code, allowing it to bypass the API when
> accessing the accessable scene-graph and run as fast compiled code.
>
Collision detection should not be scripted (IMHO).
> HOWEVER, I think the general framework for VRML should not limit it to
> mearly sharing data representing 3D physical spaces. No reason why the
> same standard suite can't be used for whiteboarding and conferencing, etc.
>