Barend Gehrels wrote:
I don't know whether convex hull should be considered as a basic polygon algorithm. It is the base of some other algorithms, e.g. the polygon diameter (furthest point pair, e.g. http://pagesperso-orange.fr/colin.barker/lpa/anti_pod.htm). But indeed it can be discussed if it is really basic. It is prescribed by OGC and implemented by most spatial databases.
I'm not sure what you mean by "basic polygon algorithm". If you could qualify that phrase I'm sure it would help myself and everyone else wondering what that means as I never knew there were "degrees" of polygon algorithms. A convex hull is by definition, simply, a set of points. The cardinality of such a set is infinite. From a geometers point of view a hull is its own structure or entity and has its own set of operations. Just because a couple of GIS applications you've work with in the past seem to have a representation of CHs that resemble a polygon doesn't mean that has to be the case. Again the concept you mentioned above regarding rotating calipers works on a set of points, these may seem like semantic differences to you if all you do is think in 2d or 3d all day long, but if you're planning on proposing a general purpose library that is expected to extend to multi dimensions and provide more than just GIS functionality then you have to start thinking about these other issues.
This is a difficult one, the one that I wanted to avoid indeed, so I omitted it from my reply. As a Boost-user, always working with and interested in geometry, this library in its current form is of no use for me. But of course, I'm part of the team of another geometry library, having most of this one and more... so why would I use this one... Therefore I feel I can't vote.
In short this is a discussion about the applicability and fitness of a polygon library, not a "mine-is-better-than-yours" debate. Arash Partow