[Gzz] Vobs and scene graphs

[B. Fallenstein]

Re the recent discussion on the jyu list about vobs and Berlin/Fresco 
(www.fresco.org). This made me think about the applicability of vobs to 
scene graph systems, again.

In a sense a VobScene represents a scene graph. A vob system could be 
implemented in which each VobScene is basically a scene graph with keys 
from the model being attached at each node.

However, this doesn't flow well with the way these graphs are frequently 
used (e.g. in Fresco): as mutable display models. This is in contrast to 
VobScenes, which are completely rebuilt every time. For example, if 
there's a window X, one must be able to say, "move this to (x,y)." This 
doesn't create a new scene graph, but modifies the existing one.

In a vob system working "just right," this would then of course 
interpolate the window to its new place. The question now is, how to 
archieve this?

Obviously two versions of the tree must be managed somehow, so that we 
can run the interpolation routines on them. The simplest possibility 
seems to be copy-to-root with immutable nodes in the scene graph: i.e., 
we create a new scene graph that shares most objects with the previous 
one, somehow. The other possibility I see is to have the nodes somehow 
store multiple versions efficiently (there are, as Tuomas continues to 
remind me, a number of efficient techniques for storing versioned data 
in memory).

In any case, this seems to be a tad difficult to add on an existing 
system like Fresco-- but I don't know.

Another comment: I don't think Fresco supports connections between two 
different branches in the scene graph, from what I read. This is very 
important to us-- and could be interesting to them? But I would not know 
how to implement that in their architecture.

- Benja