Charles Fusner wrote:

> 1. A
> 2. B
> 3. A
> 4. A
> 5. B
> 6. A

> I straddled the fence for long time on this question before
> falling gracelessly on the side of making normals more like
> the surface displacements they emulate. Even if this means

> adding a #version switch or potentially breaking every scene
> ever made for POV, (backward compatibility is a tyrrant!), but
> I just can't help liking the displacement emulating paradigm
> better than having to separately adjust bump_size. It's simply

This is exactly what I think (and even tried (and possibly failed)
to express a few months ago (I interpreted it as a bug)): Modifiying
a normal on a scaled surface is different from modifying the
surface and then computing it's modified normal.
Vanilla povray does the former, which not only is (IMHO)
counterintuitive, but even constructs normals which cannot belong
to any surface (they fail to fulfill the integrability condition).
This should be visible if a modified copy of example 5 with
"<5,1,1>" instead of "<5,1,5>"is run.

Many thanks to Nathan for creating a web page which makes
this distinction very clear (it goes far beyond the examples
which I intended to make, but never found the time to do).

Btw., even if ABAABA does not win the election, I'd wish to see
1=4, 2=5, 3=6.

Ralf

Post a reply to this message