Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!ames!pasteur!ucbvax!ucsfcgl!pixar!efo
From: efo@pixar.UUCP (efo)
Newsgroups: comp.graphics
Subject: Re: Defining a sphere with Bezier patches
Message-ID: <1681@pixar.UUCP>
Date: 2 Apr 88 01:41:08 GMT
References: <2390@saturn.ucsc.edu> <1632@pixar.UUCP> <1639@pixar.UUCP> <3183@csli.STANFORD.EDU>
Organization: Pixar -- Marin County, California
Lines: 35
Keywords: sphere, Bezier, REYES
Summary: Reasons to be cheerful (part 2)

In article <3183@csli.STANFORD.EDU>, rustcat@csli.STANFORD.EDU (Vallury Prabhakar) writes:
> In article <209@granite.dec.com> jmd@granite.UUCP (John Danskin) writes:
> 
> # Ah, but what if your modeler only supports Beziers? And, what do you mean
> # by more efficient?
Thus begins an analysis of the relative merits of modelling using Bezier
patches and surfaces of revolution.

In defense of the original questioner, I should point out that
there are plenty of reasons to want to define a sphere in terms of Bezier
patches.  For instance, as Danskin mentions, your modeller  may support
only Beziers.  Better yet, your renderer may support only Beziers.
Since the Bezier is a pretty general surface (other patches may be better,
eg, nurbs) you can create an entire renderer which knows only Beziers.
This may (or may not) make your renderer considerably simpler than one
which supports a range of other surfaces -- surfaces of revolution,
for instance, would be of little use in modelling rectilinear things
or bendy-twisty things. You can turn all manner of higher level primitives
into patches as a prepass to rendering.  This method of abstraction
may be very useful in the modelling-rendering paradigm.  However, if you're
not interested in the rendering phase a different division of labor
might be useful.

More interesting is what you might be able to do with a sphere modelled
as patches.  For instance, you might wish to use your Bezier patch sphere
as raw data to be manipulated. You might be trying to construct something
quite unlike a sphere in the end, but the sphere data is a useful starting
point.  In this case, the ability to turn a variety of things into patches
and then mung up the patches and push them together might be an interesting
way to model stuff.

For strict efficiency's sake, there are many places where Prabhakar's arguments
are the limiting ones.

Regards -
Eben Ostby