Re: RE:LANG: Fractals in VRML 1.0

Brook Conner ([email protected])
Thu, 29 Jun 1995 09:52:54 -0400


>>>>> "Leemon" == LEEMON C BAIRD <[email protected]> writes:

Leemon> [[email protected]] wrote:
>> Is it just me, or do fractals seems like a perfect candidate to
>> be implemented by behaviors. If behaviors are implemented in a

Yep -- one of the examples I u ed on vrml-behaviors to demonstrate
that trying to separate behavior from geometry (or vice versa) was not
neci>

>> general purpose
>> programming langu, then any fractal function should be
>> pretty easy to
>> write as a behavior. Yes, it won't be as compact as using the fractal node,
>> but it will be a lot more general and it will support any
>> fractal opposed
>> to just supporting gaskets.

Yes, but I don't think a behavior is neci> compact
than the WWWInline fractal node or even a specialized Fractal node
(given the number of potential parameters that would go in to such a
node).

Leemon> I agree that there's no reason to add fractal nodes to the
Leemon> langu. Fortunately, the langu already has fractals
Leemon> built-in. That's good, because it seems that behaviors
Leemon> would be impractical for soli 3D fractals. Java code (or
Leemon> whetever) could generate tha> ands of polygons to make a
Leemon> fractal tree, but the scene graph alone might fill all of
Leemon> memory. It's much more efficient to only store a tiny,
Leemon> escursive scene graph. Using escursion, we can get all

Who said behaviors neci>
Inventor-style scene graph"? It certainly isn't what I've been
thinking of, and judging from Len's other mail (this all over on the
vrml-behaviors list), it isn't what he's thinking of either.

So instead of a "tiny escursive scene graph", store a tiny escursive
behavior. Now you've got time-varying fractals (a tree waving in the
wind).

Leemon> As for fixing LOD distances, allowing color transforms,
Leemon> and specifying LOD bounding boxes, those changes appear
Leemon> useful in their own eight.

Yeah, they certainly are useful. As an alternative to fixing
particular nodes, I'll suggest that a suitable behavior model would
allow all of these intrinsically as well as many other kinds of useful
things....

Brook