There are many equivalent formulations for this problem, one of which has linear objective and convex quadratic inequality constraints. Using this formulation, LOQO was used to design the bracket shown above. It should be noted that areas in which the bracket appear to be paper thin would actually be left out of the bracket that would be build from this design---they only appear here because it is hard (i.e, inefficient) to tell VRML to leave them out. To get a better feel for the bracket, set it in motion by dragging and releasing the globe.