Support for directional data

[Renaud Detry]

Hi -

>> K-d trees seem very
>> dependant on the Euclidean metric;
> kd-tree are not linked to euclidean distances in anyway. In fact they
> have nothing to do with it.

Ok, I was definitely wrong there :-)

I'm still curious about this issue though. I'll take your advice with
the CGAL doc and other documentation. One quick question though:

> So I would like to propose that we include a distance functor as a
> parameter of the nearest neighbor calculation :-D

What do you think would be the requirements on that metric to allow
nearest neighbor queries to take advantage of the underlying Kd-tree
organization of the data?

Renaud.