I agree with Alasdair. My observations so far:
- If one wants to query a 3D scene(view) visible from a fixed viewpoint
then its much better to use z-buffer type approaches. Its much faster and
sufficient to compute distances to nearest visible parts. I believe this is
already incorporated in GIS software e.g. ArcView's 3D analyst, which uses
an OpenGL viewer window.
- There is a need to discuss what a 3D-VGA most useful for. A
straightforward extension of 2D- VGA applications would not always make
sense. I would like to differ though on the relevance of 3D space to people
behaviour. I think its relevant to characterise the 3D view from a location
because whats standing (and hanging in the air), would also effect how
people use/perceive the space. For example, my experience of Tokyo and New
York is made up of mostly three-dimensional memories/images that includes
sky scrapers, high neon billboards, bridges etc.
- On a more practial note, one application that could potentially benefit
from 3D VGA is that of CCTV installation.