"Rune" <run### [at] mobilixnetdk> wrote in message
news:3c698a7b@news.povray.org...

> Wow, this is really impressive!

I'm truly honored to know that I've impressed the great Rune! I'm really
glad you liked it!

> I don't recall in what program you make the simulation. Ultimately I hope
it
> can be implemented in POV-Ray as a patch, if that isn't what you're
already
> doing.

I'm afraid I will not be able to implement it as a patch, at least in the
short-run, because my programming abilities are not very good :)
However, I know that Apache IS trying to do a more general stand-alone
application. I'm sure everyone will love it. He has several great ideas and
I'm sure he'll be able to implement them wisely.

> Now, I know it's very ambiguous, but IMO to make the cloth simulations
> really useful for a wide variety of animations, three things are needed:
>
> * Any points on the cloth can be "control points", i.e. points that you
can
> control and attach to things in scene. For example, in this animation it
> seemed like the two upper corners were control points, or maybe the entire
> upper edge of the cloth (it was a bit difficult to see). But I hope it can
> be possible to make any point a control point. This is probably not too
> difficult; maybe your system can already do this.

Fortunately, I think it is that is not very difficult to do that. However, I
have read that it is not very wise to modify a particle's position directly,
that is, bypassing it's velocity and force. Nevertheless, that's the way I
do it, and it works... :) (more or less!!!)

> * Collision detection with any object in POV-Ray. And the objects should
be
> described in normal POV-Ray syntax. This of course is only possible if the
> cloth simulation is made as a patch. This would increase the possibilities
a
> lot!

Yes, I agree with you. The only solution would be if the program is made as
a patch. However, a general syntax for basic primitive forms can be
developed without much trouble, I think...

> * And the most difficult: Instead of the cloth always being a square
piece,
> it should be able to be any form. I don't just mean any flat form, but
> really any form. The most extreme form of this would be if the simulation
> could take any mesh as basis for the shape. I know it will be extremely
> difficult creating a system that will automatically figure out where to
> place all the strings, and just to keep track on them when there's no real
> order in the grid. However, if this or something similar is done, the
system
> could be used to simulate real clothes. Clothes, and not cloth, is in my
> opinion one of the ultimate goals for a simulation system like this.

Hmm... I don't think this is the most difficult, as I believe that the
program can be adapted to use ANY mesh, so it could really have any form.
Maybe the most difficult part is designing this shapes (like clothes) and
passing the mesh to the program, along with the information of which points
must be connected to which other points in the cloth, and their desired
distance. (Or is that exactly the difficulty you meant?)

> I don't expect you to do any of this, it's just a few long-term visions
that
> I think could be cool... ;)

Well, I share with you these ideals! Currently my worst nightmare is that I
still cannot implement a more robust method, other than Forward Euler. I
haven't been able to take that from my head for several days, so I hope that
a good solution will pop out soon.

Thanks again for your valuable comments!

Fernando.

Post a reply to this message