Re: [Fwd: Re: Introducing real 3d in Gnuplot (Was: Re: please =?utf-8?q?try=09POV-Ray_and_VRML_terminals?= )]

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 12 April 2005 01:47 am, Justace Clutter wrote:

> So, you are wanting a more feature-full 3D interface, and 
> really it is the output that you are most interested in I am sure.  

No, I think you are wrong there.  I think the real interest in a true
3D interface should be that we can support interactive 3D plotting via
OpenGL, VRML, or other front-end terminal types.  The interactive mode
of splot is relatively primitive compared to a real 3D visualization
tool.  

Pretty 3D output imaging is a separate goal, and there are other paths 
available to reach it.

To me it seems that PovRay + gnuplot is not a good match.
It's a bad idea for the same reason that adding MSOffice-style
"3D bar charts" is a bad idea. The addition of full ray-tracing to the
kind of plots gnuplot makes would add "eye-candy", but would if anything
degrade the actual information content.

Notwithstanding that, if people really want glossy 3D output images
I offer a separate package that I have maintained for many years:
	http://www.bmsc.washington.edu/raster3d/raster3d.html
Raster3D wasn't designed for this purpose, but then again neither was
PovRay.  My package renders *much* faster than PovRay, because it
"cheats" by using algorithms that are not truly ray-tracing.
This is not much of a limitation unless you really need to model higher
order properties like internal relection or refraction.
On modern hardware it is very nearly fast enough to use as a 
software rendering engine for interactive use.  But not quite -
a hardware-assisted OpenGL terminal would still be better for
interaction, though not as good for final image quality.
Raster3D also has the advantage that it is designed for use via a
pipe, and writing a gnuplot terminal driver for it would be relatively
simple (well, relative to PovRay anyhow :-).

Please note that I'm not really pushing for development of a Raster3D 
terminal type, although I do think it would be more suited than PovRay.
To return to my original point, I think the driving force for revamping
3D plotting in gnuplot should be the creation of an interactive 3D
terminal.  I would have thought OpenGL the obvious way to go, but
VRML is also a possibility.

> PovRay is a well developed application that does its job very 
> well.

I agree. But its "job" is quite different from gnuplot's.
Let's get our job description drafted before we hire anyone :-)


-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195