RE: [boost] Re: Interest for quaternions (and octonions)?

From: Reid Sweatman (borderland_at_[hidden])
Date: 2001-01-16 18:00:29

> -----Original Message-----
> From: Hubert.Holin_at_[hidden] [mailto:Hubert.Holin_at_[hidden]]
> Sent: Tuesday, January 02, 2001 4:59 PM
> To: boost_at_[hidden]
> Subject: [boost] Re: Interest for quaternions (and octonions)?
>
>
> 	That being said, what generalisation would you be interested in?

Well, that's the big question, isn't it? <g>  Certainly one that would be
compatible with rotation matrices in E3 (the space, not the conference).
But as you noted, that's hardly a fixed question.  Since both quaternions
and rotation matrices, when viewed as specifying orientation from some fixed
basis, represent constrained overspecifications, possibilities multiply.
The sign issue for quaternions is purely conventional, and most people would
assume the sign that gave the simplest-looking equations.  But then,
rotation matrices are equally conventional; as just one example, do you use
right or left matrix multiplication in your code?  Being an OpenGL
programmer, I have a preference there, but much of the world wouldn't agree
with me.  And there are plenty of other ways of overspecifying rotations,
even using matrices alone, without going to higher-order objects.

I suppose a better question would be:  what is most doable in your existing
codebase?  While not sacrificing efficiency, of course.

Reid Sweatman
Software Engineer