The specs state: "you can position a camera by placing a transformation
node before it in the scene graph".
So why doesn't it work on the 2 browsers that I tried? Can someone provide
an example of a non-trivial transformation acting on a camera? Can the
camera be nested arbitrarily deep in separator nodes?
Ideally I would like to export the view-to-world transform then the camera.
That would automatically put the camera in the right place in the world
(with some small fiddling for focal distance etc). This just doesn't appear
to work.
Currently I export the world-to-view transformation immediately after the
camera in order to transform every other object into view space. This works
fine, and is quite a nice way of doing it (assuming browsers aren't stupid
about how they handle the transform stack).
However, I have been told that this makes the navigation system system
of some browsers do strange things, probably because Y isn't up any more.
This then is a serious flaw in the browsers. We are living in 3D here.
Get out of the stupid 2D mind set and start thinking that any direction
is an equally valid up. Because it is. If I am flying a plane I have
every right to fly upside down or sideways. If I'm in a spaceship then there
is no up or down at all. There is no need for VRML to be stuck on the
earth. Of course this may be inflamatory speculation. ;-) There are good
psychological esasons for keeping things on the ground.
Steve.