Re: [Plib-users] DAG issues

[Steve B. <sjb...@ai...>]

Rahul Choudhury wrote:
> 
> Steve:
> 
> Thanks for the clarification. I noticed that any node
> thats derived from ssgEntity can have multiple
> parents. Doesn't it mean that a scene graph
> represented by SSG is not necessarily a Directed
> Acyclic Graph ?

Yep. That's a design feature.

For example, in my Tux game, there are lots of red, green,
silver and gold herring that spin around.  Each level has
dozens of herring - and I wouldn't want to have to compute
a matrix for each of them.  However, since a single spinning
herring model can have multiple parents, I can have as many
spinning herring as I want - although they are all simple
instances of the same basic herring.

> What are the implications of this ?

Well, it makes reading and writing files "interesting".

I don't imagine there will generally be many cases where
positions of things like those herring would be held in
a single model file.  My games load basic objects from
standard file formats - but how they are arranged to form
a total scene is something that I would use a custom 'level'
file format that could associate behaviours and special effects
with those models.

That setup makes it perfectly natural to have multiple copies
of an object.
 
> I thought we only need geometry nodes (ssgLeafs) to
> have multiple parents for instancing purpose. Please
> let me know your thoughts. (any one who is interested)

No - I think it's useful (and natural) to have 'instancing'
at any level in the tree - it's not hard to do and it has
significant benefits.

-- 
Steve Baker                  http://web2.airmail.net/sjbaker1
sjb...@ai... (home)  http://www.woodsoup.org/~sbaker
sj...@ht...      (work)