^{This issue is triggered by this forum post}

Right now, the computation of the screen space error from the geometric error and the distance is linear, as in sse = geometricError / distance (with some more or less constant factors sprinkled in).

It could be a reasonable demand to use higher detail for things that are closer, taking the distance into account non-linearly. For example, sse = geometricError / pow(distance, someExponent), with someExponent being 1.0 by default, or 0.9 to give a higher detail to elements that are closer to the viewer.

_{This could be implemented trivially ("Just add someExponent to TilesetOptions and pass it along..."), but of course, if something like this is implemented, it should be offered in a more generic form. As such, this issue is only one special instance of all the thoughts and scenarios that I compressed into #126 , but to have an overview of what could be possible with a few abstractions for the key concepts like "culling", "load priorization" etc., I'll mention it dedicatedly here.}