I absolutely agree with this need. In generall it is possible to want
other graphics to be screen based (Cursors or Symbol
Graphics). Perhaps one could think of annotation text as a 3-D point
connected to a screen space text object. (This way font sizes can also
be specified in screen space, maybe as a function of scale like a LOD
node.)
Currently we are implimenting this functionality with callbacks in
Inventor. Something obviously not transportable to VRML.
-- for(;;); /* eta */