Re: [boost] Standardization of quaternion, octonion, and special function libs?

From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2002-05-24 09:14:57

On Friday 24 May 2002 15:32, Beman Dawes wrote:
> Yesterday I had lunch with Hubert Holin, who was in the US to present at a
> conference.
>
> We had some discussion about him submitting a proposal to the C++ committee
> to standardize his Boost quaternion, octonion, and special function
> libraries.
>
> One of the key issues is the level of interest from the numerics
> community.  There just isn't any way this (or any other numerics proposal)
> will be accepted by the committee without support from the numerics
> community.

This is actually a structural problem. Engineers in numerics focus mostly on 
the numerics and mostly don't care for the best possible implementation from 
a software engineering point of view. Therefor, few numerics guys or others 
using a lot of numerics in their developments, generally don't know about 
boost or other great libraries/techniques/papers out 
(for example, if you know how many people buy the RogueWave matrix library 
but on the other hand how few are using ublas or MTL or ...).

And this gives us a dead-lock situation. Many people programming numerics 
complain that e.g. C++ does not provide matrices but fortran-90 does. 
(resulting in computer-science departments using either fortran-90 or Java). 
But OTOH, as you explain, C++ will not provide these features if there are 
not many users out there.

Thus the commitee should, when deciding upon supporting features which are 
not purely in the software-engineering domain, not take the number of current 
users of the specific feature into account but the number of people that 
actually need that functionality.