Quote:

in a7 the fourth component of the tangent contains the handedness of the binormal. previously this information wasn't available and so mirrored uvs were problematic.




I have used the <cross(t, n) * t.w> and it has solved a couple of ongoing headaches I have had.