Hi Akos,

> as one of the original implementors of the NURBS support, I 
> feel I should 
> clarify some of our design decisions. :-)

Thanks. It's always good to hear from the real experts :-)

> We abandoned the sewing approach mainly because no matter 
> what methods we 
> tried (and we tried some) it never worked well for all 
> models. Usually it 
> handled some models quite well and didn't work at all for 
> other models. I 
> know some people dislike the Fat Borders approach, but it 
> works reasonably 
> well for rendering. It's also used in our GPU NURBS framework 
> which seems 
> to be quite popular, despite not even having an explicit mesh. :-)

I could not find any documentation for this. Can you please point me to
something that I can take a look at and play with? Is this part of the
CVS tree? The only thing I could find on google was this 2002 paper from
U Bonn which talks about the Sewing algorithm in OpenSG Plus. Nothing
about Fat Borders or the GPU NURBS stuff. 

> As for collision detection, I think it's actually possible to use the 
> OpenSG generated meshes even though they're not crackfree, 
> because you 
> still get guaranteed geometric precision in model space (which is not 
> trivial to achieve).

That is good to hear. 

> AFAIK topological reconstruction of arbitrary NURBS models is 
> still an 
> open question. I'm not an Optimizer expert by any means so 
> feel free to 
> correct me here but I don't think Optimizer's one pass/two pass 
> reconstruction methods work that well in bad cases (e.g. non-manifold 
> topology). Actually the description of their two pass method 
> sounds rather 
> similar to our seam-graph approach, which I know didn't work 
> in all cases 
> (and we spent a _lot_ of effort on that one).
> 
> Also AFAICS Optimizer more or less tessellates into a grid 
> which may or 
> may not be what you want, plus I really don't think it 
> guarantees you a 
> completely gapfree mesh (even inside one topology).
> 
> Anyway, if you want to tackle writing your own sewing-method, 
> the best 
> approach would probably be writing some kind of 
> "sewing-Action". However, 
> even if you have some kind of topological information this is 
> far from 
> trivial, consider e.g. two adjacent surfaces which are both 
> trimmed on the 
> joint side with differently parameterized curves: you're 
> basically out of 
> luck... (and this is not an uncommon scenario in industrial 
> models that 
> I've seen so far.)

Thanks a lot for your inputs. I guess, I will try what is available
today in OpenSG and then see if that is 'good enough' for our
application/data sets or if we need to improve on that. If it turns out
to be the later, I will make sure that I discuss with the rest of the
folks on the list before diving into it directly.

- Praveen


Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to