I get that the algorithm for pitch shifting used in the video above simulates Doppler effect, the formula of which being : f_out = f_in ( (v_sound + v_observer) / (v_sound + v_source) ).
It is simulated by putting a phasor into the delay line, which causes linearly increasing delay time. The frequency of the phasor (say f_phasor) is v_observer / v_sound. So the formula for f_out becomes f_out = f_in ( 1 - f_phasor ).
Here is what I don't get : apparently increasing delay time is analogous to 'moving observer' in the Doppler effect. Why not 'moving source'? What is the difference? Aren't both cases the same in that the signal arrive-time gets more delayed?