http://www.aip.org/history/einstein/essay-einstein-relativity.htm John Stachel: "But here he ran into the most blatant-seeming contradiction, which I mentioned earlier when first discussing the two principles. As noted then, the Maxwell-Lorentz equations imply that there exists (at least) one inertial frame in which the speed of light is a constant regardless of the motion of the light source. Einstein's version of the relativity principle (minus the ether) requires that, if this is true for one inertial frame, it must be true for all inertial frames. But this seems to be nonsense. How can it happen that the speed of light relative to an observer cannot be increased or decreased if that observer moves towards or away from a light beam? Einstein states that he wrestled with this problem over a lengthy period of time, to the point of despair. We have no details of this struggle, unfortunately. Finally, after a day spent wrestling once more with the problem in the company of his friend and patent office colleague Michele Besso, the only person thanked in the 1905 SRT paper, there came a moment of crucial insight. In all of his struggles with the emission theory as well as with Lorentz's theory, he had been assuming that the ordinary Newtonian law of addition of velocities was unproblematic. It is this law of addition of velocities that allows one to "prove" that, if the velocity of light is constant with respect to one inertial frame, it cannot be constant with respect to any other inertial frame moving with respect to the first. It suddenly dawned on Einstein that this "obvious" law was based on certain assumptions about the nature of time always tacitly made."
The law of addition of velocities cannot be rejected so easily - it is a logical consequence of the Doppler frequency shift. If the observer starts moving towards the wave source with speed v, then, for all waves (light waves included), the frequency as measured by the observer shifts from f=c/L to f'=(c+v)/L=f(1+v/c), where L is the wavelength:
http://www.hep.man.ac.uk/u/roger/PHYS10302/lecture18.pdf Roger Barlow, Professor of Particle Physics: "The Doppler effect - changes in frequencies when sources or observers are in motion - is familiar to anyone who has stood at the roadside and watched (and listened) to the cars go by. It applies to all types of wave, not just sound. (...) Moving Observer. Now suppose the source is fixed but the observer is moving towards the source, with speed v. In time t, ct/lambda waves pass a fixed point. A moving point adds another vt/lambda. So f'=(c+v)/lambda."
http://rockpile.phys.virginia.edu/mod04/mod34.pdf Paul Fendley: "Now let's see what this does to the frequency of the light. We know that even without special relativity, observers moving at different velocities measure different frequencies. (This is the reason the pitch of an ambulance changes as it passes you it doesn't change if you're on the ambulance). This is called the Doppler shift, and for small relative velocity v it is easy to show that the frequency shifts from f to f(1+v/c) (it goes up heading toward you, down away from you). There are relativistic corrections, but these are negligible here."
The only reasonable conclusion is that, for all waves (light waves included), the speed of the waves relative to the observer shifts from c=fL to c'=f'L=c+v, in violation of special relativity:
http://www.einstein-online.info/spotlights/doppler Albert Einstein Institute: "The frequency of a wave-like signal - such as sound or light - depends on the movement of the sender and of the receiver. This is known as the Doppler effect. (...) In the above paragraphs, we have only considered moving sources. In fact, a closer look at cases where it is the receiver that is in motion will show that this kind of motion leads to a very similar kind of Doppler effect. Here is an animation of the receiver moving towards the source: (...) By observing the two indicator lights, you can see for yourself that, once more, there is a blue-shift - the pulse frequency measured at the receiver is somewhat higher than the frequency with which the pulses are sent out. This time, the distances between subsequent pulses are not affected, but still there is a frequency shift: As the receiver moves towards each pulse, the time until pulse and receiver meet up is shortened. In this particular animation, which has the receiver moving towards the source at one third the speed of the pulses themselves, four pulses are received in the time it takes the source to emit three pulses."
http://a-levelphysicstutor.com/wav-doppler.php "vO is the velocity of an observer moving towards the source. This velocity is independent of the motion of the source. Hence, the velocity of waves relative to the observer is c + vO. (...) The motion of an observer does not alter the wavelength. The increase in frequency is a result of the observer encountering more wavelengths in a given time."
http://www.radartutorial.eu/11.coherent/co06.fr.html "L'effet Doppler est le décalage de fréquence d'une onde acoustique ou électromagnétique entre la mesure à l'émission et la mesure à la réception lorsque la distance entre l'émetteur et le récepteur varie au cours du temps. (...) Pour comprendre ce phénomène, il s'agit de penser à une onde à une fréquence donnée qui est émise vers un observateur en mouvement, ou vis-versa. LA LONGUEUR D'ONDE DU SIGNAL EST CONSTANTE mais si l'observateur se rapproche de la source, il se déplace vers les fronts d'ondes successifs et perçoit donc plus d'ondes par seconde que s'il était resté stationnaire, donc une augmentation de la fréquence. De la même manière, s'il s'éloigne de la source, les fronts d'onde l'atteindront avec un retard qui dépend de sa vitesse d'éloignement, donc une diminution de la fréquence. Dans le cas sonore, cela se traduit par un son plus aigu lors d'un rapprochement de la source et un son plus grave en s'éloignant de celle-ci. Dans le domaine de la lumière visible, on parle de décalage vers le bleu pour un rapprochement et vers le rouge dans le cas d'éloignement en se référant au spectre lumineux. La même chose s'applique à toutes les gammes d'ondes électromagnétiques dont les ondes utilisées par les radars."
http://physics.bu.edu/~redner/211-sp06/class19/class19_doppler.html "The Doppler effect is the shift in frequency of a wave that occurs when the wave source, or the detector of the wave, is moving. Applications of the Doppler effect range from medical tests using ultrasound to radar detectors and astronomy (with electromagnetic waves). (...) We will focus on sound waves in describing the Doppler effect, but it works for other waves too. (...) Let's say you, the observer, now move toward the source with velocity vO. You encounter more waves per unit time than you did before. Relative to you, the waves travel at a higher speed: v'=v+vO. The frequency of the waves you detect is higher, and is given by: f'=v'/(lambda)=(v+vO)/(lambda)."
http://www.usna.edu/Users/physics/mungan/Scholarship/DopplerEffect.pdf "Consider the case where the observer moves toward the source. In this case, the observer is rushing head-long into the wavefronts, so that we expect v'>v. In fact, the wave speed is simply increased by the observer speed, as we can see by jumping into the observer's frame of reference. Thus, v'=v+v_o=v(1+v_o/v). Finally, the frequency must increase by exactly the same factor as the wave speed increased, in order to ensure that L'=L -> v'/f'=v/f. Putting everything together, we thus have: OBSERVER MOVING TOWARD SOURCE: L'=L; f'=f(1+v_o/v); v'=v+v_o."
http://www.cmmp.ucl.ac.uk/~ahh/teaching/1B24n/lect19.pdf Tony Harker, University College London: "The Doppler Effect: Moving sources and receivers. The phenomena which occur when a source of sound is in motion are well known. The example which is usually cited is the change in pitch of the engine of a moving vehicle as it approaches. In our treatment we shall not specify the type of wave motion involved, and our results will be applicable to sound or to light. (...) Now suppose that the observer is moving with a velocity Vo away from the source. (...) If the observer moves with a speed Vo away from the source (...), then in a time t the number of waves which reach the observer are those in a distance (c-Vo)t, so the number of waves observed is (c-Vo)t/lambda, giving an observed frequency f'=f(1-Vo/c) when the observer is moving away from the source at a speed Vo."