I like Sneftel''s approach, and it should work well as long as your model is relatively smooth and if you deal with each separate mesh object individually.

Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.

The only problem is drawing a line passing from a point which
has the components normal to the surface normal
that''s to say a*l+b*m+c*n = 0 and this is not an easy task
it works if the triangle is aligned with some axis