Coverage or Distance for swithing threshold?

[Takahiro]

Currently the extension allows to define screen coverage to determine the LOD switching threashold. Another option may be distance from the camera.

The both have Pros and Cons. Which one should be used? Or is there any more generic parameter?

Screen coverage

• Pros: Better quality. For example if an application scales up a distant model and the model will be more visible, higher LOD level will be selected.
• Cons: Harder to compute. Because the computing needs some dependencies eg. camera info, world scale, LOD and descendant geometry (bounding box) sizes. And screen coverage of a model needs to be dynamically re-computed if camera parameter or model's world matrix is updated.

Distance Pros and Cons

• Pros: Easier to compute. No dependencies needed unlike screen coverage computing. And no dynamically recomputing is needed.
• Cons: Worse quality. For example even if an application scales up a distant model and the model will be more visible, lower LOD level will still be selected. Then it will end up that the model will be noticable with low quality model.

I may prefer distance because of easiness? But not sure if all the major engines support distance based threashold. (I feel screen coverage based threashold is more standard now?)