[geeks] ahem

[Joshua D Boyd]

On Sat, Jun 22, 2002 at 01:23:04AM -0400, Kurt Huhn wrote:
> > Did you ever try Ayam?
> > http://www.ayam3d.org/
> > 
> > They take their NURBS seriously.  It is based on MOPS, which I've used
> > and liked better than most NURBS modellers.
> > 
> 
> I tried it very cursorily, got impatient with the interface, and
> stopped.  However, I'll probably come back to it.  
> 
> I'm fucking pissed off at RealSoft3d right now.  The interface is *ass
> backwards* - while the render seem top-notch, making something really is
> an excersise in patience.  Something simple, like gold ring, will take
> me a total of 90 seconds from start to render in Rhino with the RhinoMan
> BMRT interface - in RealSoft it took me 90 *minutes*.  Some of that can
> be attributed to an unfamiliar interface, but mostly it's just because
> you can't manipulate things like you can in Rhino.  RealSoft3d truly has
> a bad interface - not just the buttons and clicky things (that's fully
> customizeable) but the *manner* with which you create and manipulate
> NURBS, surfaces, and solids is friggen wierd and complicated.

Would you care do describe the actual workflow in RealSoft3d, and your
ideal workflow?

See, my ideal work flow for NURBS would be (depending on object being
modeled) to either start with a primitive shape (like a sphere), then
add more curves to the areas I want to drag out into more complex
shapes.  I suspect this is a "bad" way to try to deal with NURBS
though (most programs make it fairly hard to do it this way).

My other method for dealing with it was to outline the figure to be
created, then try to loft surfaces out of the outlines.

I seem to remeber that taking splines and creating surfaces from them
via lofts, blends, trims, etc, was the Rhino prefered way of doing things.
But really, the complexity issues always got to me (to add localized
complexity, like to a facial feature, you must raise complexity of the
entire model, etc,etc,etc).  Of course, theoretically, you are
supposed to deal with by splitting your figure into multiple models,
but then, making it look like one model again is non-trivial.

NURBS were always really good to me for objects that were simple lofts
though.

On a note related to not just using them in a cad/modelling program,
of all the manners for curved surfaces I've looked at mathematically,
from implicit surfaces, CSG, subdivision, bezier patches, and more
that are uncommon, NURBS are the hardest and most complicated to
grok.  Heck, in my experience in 2D curves, B-Splines (rational or
not, uniform or not) are just about the hardest to deal with (though
that may be a matter of bad presentation from professors).

Subdivision surfaces are also pretty tricky, but thankfully, the tools
for working on them are virtually the same as for working on polygons.
I'm still trying to find an easy explaination of the math behind
subdividing.  So far, it seems one key point is to store models in
winged edge trees instead of vertex lists.

-- 
Joshua D. Boyd