Re: Geometry Chapter of the manual

[John W. Eaton]

On 24-Jul-2007, David Bateman wrote:

| Yes adding a dependency means that Octave becomes progressively harder
| to build. However, it also means the addition of new functionality with
| the minimum work on the part of the Octave developers. There is a
| compromise to achieve hear and I think its normal that the addition of
| each new dependency to the core of Octave should be considered on a case
| by case basis with some discussion of interested parties.
| 
| So in particular, in your opinion, does the inclusion of the convhull,
| convhulln, delaunay, delaunay3, delaunayn, griddata, tsearch, voronoi
| and voronoin function justify the inclusion of the dependency on QHull?
| Are these functions of sufficiently large value to accept the cost of a
| harder build of Octave?

Since all these functions are core Matlab functions, I think people
will expect Octave to have them.  It seems the best way to provide
them is by using the qhull library, so I don't see a good way around
the dependency.  Since this is a separate library, I think we should
treat it the same as others like fftw, glpk, pcre, ufsparse, etc.  It
doesn't make sense to include all of these external packages in Octave
itself.

jwe