Exactly! To find the size of the bounding box you have to transform
all of the bounding box coordinates into screen coords... Thats a
lot of calculations for wimpy machines. The distance based method
makes a whole lot more sense.
I'm assuming that the 'World-space distances' are the distance from
the eyepoint to the center of the LOD.
Should we keep the old LevelofDetail in, or just scrap it? I can
think of at least one example of where it would still be useful.
Assume that I have a model of the Statue of Liberty. This, of
course, is a very tall and skinny model. Assume that the model is
200m tall, 20m wide and 20m deep and the bottom is centered at 0,0,0
Assume that I have the following LOD (and that y is up)
LOD {
ranges [ 1000 500 100 30]
center 0 100 0
Separator { little bitty model }
Separator { better model }
Separator { shiny specular super dazzling model }
Separator { More lifelike than the real thing model }
}
For the most part LOD duplicates the behavior of LevelOfDetail, but
when you get really close a big difference happens:
If I move my eye to the center of the model (i.e. close to 0,100,0)
then I get the ultra-lifelike model.
If I move my eye to the top of the model I only get the 'better
model'..
This of course is just one special case. I think with some tweaking
(perhaps by dividing models into component sphere-like areas) we could
approximate the bounding boxes of 'LevelofDetail'
------------------
My opinion is that we go with LOD. But then again, I have my reasons.
I haven't gotten around to implementing LevelofDetail yet. LOD seems
much simpler to implement.
Plus, I'm one of those Browser geeks :-)
-- | Paul Lindner | [email protected] | Slipping into madness | | Distributed Computing Services | is good for the sake | Gophermaster | University of Minnesota | of comparison. ///// / / / /////// / / / / / / / / //// / / / / / / / /