Why not use a sophisticated library as CGAL or some other?
CGAL is substantially QPL licensed - dragging us back to the dark-ages of licensing IMHO.
It is entirely preferable to avoid CGAL on that basis.
Obviously we ARE in the dark-ages of licensing, GPL seems to be the problem here.
can this problem be solved like this?
Seriously - relying on QPL code is just a disaster area. It couples us to what will become the next failed business model (I suspect). Furthermore, I (for one) would really like to incrementally enable an LGPL code-base here to allow people to re-use and embed the code left and right; moving to use and depend on a "separate patches" "non-commercial use only" type thing is hyper annoying.
We spent a good while ridding ourselves of 'gpc' for it's similarly dumb license; it'd be good not to get these problems back.
Clearly re-using algorithms inspired from CGAL makes sense (they are mostly commodity ones anyway); and IIRC thaytan was looking at some of that.
I don't see your point, we would have no "separate patches" or "non-commercial use only" issue because we would not be changing CGAL, just linking
other suggestion: boost (http://www.boost.org/doc/libs/1_47_0/libs/geometry/doc/html/index.html) which is "GPL compatible"
Yep - boost makes a lot more sense :-) And of course, we'd want to ship repsnapper to lots of users, sometimes in a commercial context and/or statically linked etc.
now using Clipper http://angusj.com/delphi/clipper.php