On Mon, 10 Feb 2003 14:38:44 +0100, Christoph Hormann
<chr### [at] gmxde> wrote:



> I cannot even imagine how much it took to program this into
> a raytracer (a great job !!),
I mean this, a complement well deserved.

>Feel free to think whatever you want to think but i see no argument in
>your writing that would support this extremely general statement.
Ok, might be too general, see more nuance below

>> You can do these kind of things much faster, more accurate and without
>> the risk of crashing your raytracer with external modellers and
>> simulators.
>
>This strongly indicates that you don't really know what you are talking
>about.  
Statements like this will not really engage people in a
discussion..... Plenty of managers in high tech company's pay loads of
money to send their engineers to courses in order to train them not to
make this kind of statements. Trust me on this.

>First of all the mechsim patch is not any more likely to crash
>POV-Ray than any other experimental feature.  
True, but I am not comparing with other POV features, I am comparing
with mechanical simulation systems which have, as you know, lots of
mathematical appoximations and in this case "spring & damper" systems
which are inherently unstable when used without extreme caution.
These should not be part of a renderer which should just render any
datafile whe feed it.

>And then i have serious
>doubts that Aero is substantially faster or more accurate simulating the
>same system than the mechsim patch.
From an end-user perspective it certainly was :), it really took
me just a couple of minutes. You give the reason below; I did have the
need the build objects from point masses and to apply the "correct"
stiffness factors. Wooden domino blocks do not bend or deform
when they fall. That is what triggered me in the first place.

Could we compare ? i.e. create a sample scene, a sort of a skyvase
for mechsim and compare on speed an accuracy ? See where the limits
are ?

>What probably makes Aero more appealing to you is the ability to simulate
>the movement of a few basic shapes like boxes and cylinders without the
>necessity to build them from  point masses like in mechsim.  
Look what it is used for by the other users; they use it for modelling
the behaviour of in-flexible masses, like domino blocks, or balls
falling down a wall filled with obstacles.  

Yes I like the simplification approach. Simulation simplified models
correctly gives i.m.h.o. better results than larger, unstable, models.
But maybe recent models and CPU power are catching up on me.

>have a look at the Aero documentation you will recognize that there is a
>serious number of tricks necessary to make this work.  This might produce
>reasonable results for falling domino stones but will break into pieces as
>soon as you try a simulation of more complex shapes and deformable bodies.
I know the limitations of aero and the tricks. Deformable bodies are
the holy grail of course. Can you do this ?  Can you make a cube which
falls on a corner and actually deforms permanently ?



>BTW animations posted in several pieces will be difficult to view for a
>lot of people.  
>Christoph

Post a reply to this message