My opinion is that Stokes made the most important error when derived an effect of ether wind in the transverse arm in MMX. This effect does not exist, since the mirrors in the interferometer do NOT cause the vector sum of ether wind and wave velocity to be ortogonal to mirrors. Instead wave velocity only is ortogonal to mirrors. Therefore NO effect in the transverse arm in MMX. Atoms in solids control separations by generating fields that propagate with light speed. Therefore, reduction in 2-way light speed is COMPENSATED by length contraction. This means that MMX cannot detect an ether wind.
I have described this on my blog.