> I would disagree, I believe your implementation has a series of flaws that
> render it less robust. Your implementation can lead to false positives on
> intersections that should not be found.

Exactly. This is how I workaround numerical imprecisions. And you are doing 
the same in your patch. We must just not return negative on the place of 
positive.

I accept your authority over what is included in OSG and what is not. I just 
offer you that I might work on the solution in the future if the problem, by a 
chance, comes back on some very curious and strange models. I have done some 
assembly FPU programming, so I touched the metal, and I made three year 
research on data consitency in distributed systems where one of the problems 
was non-determinism and imprecision of FPUs. So, I think, I might have few 
words to the topic. But until we find a problematic model, why to use the 
complicated code that is probably too conservative and very poorly readable? I 
think that your decision of simple solution is based on solid ground.

John
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org

Reply via email to