There has been a lot of discussion about projects that sound way too ambitious for the time available for a GSoC. Agree, it shold be concised. It would be useful to have even a 2D library in a reviewable state.
If there is any time left over, other 3, 4 and other fixed N can be explored or tackled?
What we, as geometry developers, at least need: - a fixed size matrix where N is a template parameter - should be able to work with large number types (optional, via interface or other, see other discussion on mailing list about GMP) - matrix determinant - matrix multiplication - matrix inversion Most is for transformations. We're now using ublas but indeed a fixed size matrix would me more convenient and probably more efficient. For 2D geometry a matrix size 3 is necessary.
But maybe somebody can argue that we
shouldn't have a fixed-size matrix class? +1
Regards, Barend