[vos-d] VOS requirements

[Karsten Otto]

Am 26.01.2007 um 13:35 schrieb Len Bullard:

> Some interesting papers on associating semantic metadata with X3D  
> scene
> graphs have been published at last year's Web3D Conference and more  
> at this
> year (I am a reviewer but cannot comment more than that).  So it is  
> doable
> and has been done.
>
Oh, I never doubted that. I have read the paper you mentioned, and I  
am most certainly interested in the new ones; I'll have a look once  
the proceedings are available. Thanks for the pointer!

> For VRML, it came down to requirements.  At the time of the  
> specification,
> it was already too heavy for the web infrastructure.  Now, it is about
> right.  Add more metadata then was not technically smart; now it is
> inconvenient.
>
Agreed. I followed the design process (at least the part visible to  
the public), and it seemed very much a "we want this now" situation,  
calling for quick adoption of *something*. Dropping dynamics in VRML1  
and postponing multi-user-capability until much much later were the  
consequences. Too bad...

> What you want is an object-oriented programming language, not a  
> declarative
> scene graph with scripts.  I understand that.  Had the VRML2 designers
> chosen that design, the Microsoft submission would have won.   
> Possibly that
> would been better for what you want.
>
Well, maybe. I am not sure an OO language is a good way to express 3D  
shapes. After all, most code in scene graph implementations is  
involved in mainitaining and manipulating the graph structure; code  
that actually *does* something is pretty rare (transformations come  
to mind). The interesting parts all happen in the renderer. Pure data  
structures or a file format seem a much better solution then.

You can argue the same for semantics; use some generic file format  
like RDF/XML, the interesting stuff all happens in the application  
interpreting the RDFS/OWL/DAML contained in it.

But I always understood VOS to be primarily a distributed programming  
model... how do these things fit?

> len
Regards,
Karsten Otto