Nieminen Mika wrote:

>   People really don't know what is a bug.
>   It's very annoying that people keep sending "I don't like the way this
> works"-messages as bug reports.
>   Bugs are programming mistakes. Something should work this way but it
> doesn't because the programmer made a mistake while typing the code. Bugs

Just a question about this topic:
Recently, I found out that the "bumps" feature when scaled 
anisotropically does not behave like real bumps. I wrote this
into povray.bugreports and got a few answers from people
who IMHO were not extremely familiar with the mathematical
aspects involved (which included differential geometry).

What would be the proper action in this case?
Flooding the bugreports group with more text is
probably a bad idea. Since I cannot read the mind of
the authors, I cannot tell wether this is a program bug,
a math error, or simply a problem which nobody cared
about yet.

Ralf

PS: The problem in question is the following: Transformed
bumps are computed by evaluating the bump function at some
other place in space, then using it on the place it is intended
for. Since real bumps on things (or other normal modifications)
involve computing the derivative, we get in trouble if the
transformation does not commute with the derivative, which
becomes relevant in the anisotropic case.

It can be visualized as follows: Make a xy-plane with bumps,
and have the lightsource at the same place as the camera 
(e.g. at <0,0,-10>).
Then the bumpy plane will have a circular region where
bright spots are visible. If the bumps are elongated
(i.e. longer in the x- than in the y-direction), then
this region should be deformed into an ellipse, which is
elongated in the *y*-direction (This can be verified by
making a real 3d-surface, e.g. using a heightfield).
Povray still shows a circular region.

Post a reply to this message