[x3d-public] Normal node precision

[vmarchetti at kshell.com]

Since the Normal node data is defined as MFVec3f type, which is vectors based on "single-precision floating points", and in section 5.3.5 https://www.web3d.org/documents/specifications/19775-1/V3.3/index.html , single precision is described as requiring at least 6 (decimal) digits of precision. So a sensible specification for the normal vector data would be

abs( sqrt( n[0]*n[0] + n[1]*n[1] + n[2]*n[2]) - 1.0) <= 1.0e-6  , must be satisfied for each normal vector n with components n[i]

I judge it would be useful to include this in the specification. I don't judge it would be appropriate to specify what browsers should do for non-normalized normal vectors. This is the classic problem for software developers about hardest engineering factor, and that is users and the the mistakes they make -- and by users I mean content authors who don't properly prepare their models. Individual browsers need to make the tradeoff between accepting bad input and providing quality output; given that this tradeoff also depends on the target market.



> On May 6, 2020, at 9:49 AM, Andreas Plesch <andreasplesch at gmail.com> wrote:
> 
> I guess there are no immediate opinions on the requirement that normals need to be normalized to a length of 1.0.
> 
> How should a x3d browser process a normal with a length of 1.5 ?
> 
> A vote ? on:
> 
> a) give up. Invalidate the X3D. But what about a length of 1.01 ?
> b) use as is but proceed in the processing of shading as if it is of length 1.0. May lead to rendering artefacts but I think this is the expected behaviour.
> c) normalize the normal to 1.0 internally. I guess that is what the requirement tries to avoid browsers have to do but browser may already do it anyways for sanity reasons.
> 
> I pick b).
> 
> Cheers, -Andreas
> 
> 
> 
> On Tue, Apr 28, 2020 at 12:53 PM Andreas Plesch <andreasplesch at gmail.com <mailto:andreasplesch at gmail.com>> wrote:
> Looking at improving the pythonocc x3d generation, we came about the
> requirement in X3D that normals need to be normalized to unit length:
> 
> https://www.web3d.org/documents/specifications/19775-1/V3.3/Part01/components/rendering.html#Normals <https://www.web3d.org/documents/specifications/19775-1/V3.3/Part01/components/rendering.html#Normals>
> 
> However, the spec. is silent on how precisely that requirement needs
> to be met. Is it required to be as precise as floats allow ? As an
> example, consider
> 
> (0.9, 0.43588989435, 0)
> 
> This vector has unit length given float precision limits.
> 
> Would
> 
> (0.9, 0.43589, 0)
> 
> still be considered legal although it has a length of 1.00000004605 ?
> 
> Does the X3D validator check for this requirement ?
> 
> Thanks, Andreas
> -- 
> Andreas Plesch
> Waltham, MA 02453
> 
> 
> -- 
> Andreas Plesch
> Waltham, MA 02453
> _______________________________________________
> x3d-public mailing list
> x3d-public at web3d.org
> http://web3d.org/mailman/listinfo/x3d-public_web3d.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://web3d.org/pipermail/x3d-public_web3d.org/attachments/20200506/08c29aab/attachment.html>