How is it possible for an observer to measure the speed of light as it moves AWAY from him? I often come across statements such as this one, by N. David Mermin, a well-known physicist:

"Let P be a valid procedure for carrying out the time and distance
  measurements that allow one to determine the speed of an object in a
  given inertial frame. Let Bob, carrying out the procedure P in the
  frame of reference of a space station, measure the speed of a pulse of
  light as it zooms off into space."

In this thought-experiment, how can we simply assume in advance that Bob has such a "valid procedure" ? Doesn't this have to be demonstrated, not merely asserted?    More specifically, how could any observer measure the speed of light as it moves away from him or her?  Wouldn't the light photons have to interact with a measuring apparatus that the observer has at his disposal? - not away from him, but with him?
 A: Bob could set up detectors ahead of time that are in his frame of reference that measure the time when the light reaches them, and then later look at the results.  We can assume that Bob has done any amount of preparation, since it's just a thought experiment after all.
A: Wave frequencies change, depending whether the source of the wave moves away from or toward you.  This does not affect the speed of the wave.
As the source of a light pulse generated in your own inertial frame is moving neither toward nor away from you, the frequency of that light pulse should appear the same at all points in its journey as when it left your generator, provided that objects which reflect the light back to you are stationary with regard to your inertial frame.
As the light pulse moves away from you, it may leave traces of its passage on very dilute dust particles (for example) that are stationary with regard to your inertial frame.  Light reflecting back to you from relatively stationary dust should have the frequency of the original pulse.  Comparing incoming frequency with outgoing frequency would be a way to make certain that you are not moving with or against the light pulse.
The time between generation, reflection, and return of the original pulse can provide a way to measure the speed of the pulse, provided that distance to the point of reflection is known, and provided that the reflector is either stationary, or moving at a known velocity in a known direction relative to your inertial frame.
A: It is impossible to detect light in a vacuum, that is moving away from the detector.  
A: Setup-1
Place N stationary mirrors on the corners of a polygon with equal sides such that the pulse will loop indefinitely unless it is observed by some measurement device or any other obstacle. Take a space ship which will orbit around the center of the mirrors. As you orbit, you will pass by each mirror at equal time intervals depending on on the speed of the ship. 
As you just by pass by a mirror, send a pulse to the next mirror and continue the trip as the pulse starts its looping. Depending on the speed of the ship and N and relative direction of pulse and the ship, you can arrange things such that the ship and pulse will encounter at the Mth mirror where you would like to measure the time (locally) and try to detect the pulse  and confirm that the pulse is really there locally in that Mth mirror. Do the experiment for different M values in both counter-clock or clock wise directions. If you fail to meet the pulse for any calculated M value (by changing the speed of ship) in any orbit direction, you would start suspecting that speed of light might not be the same for away or towards direction.
Note that the ship is actually rotating. To avoid its complication, more mirrors can be put on the sides of the polygon and M values can be arranged to correspond mirrors on one experiment side of the polygon. So only the light makes direction change.
Actually whole idea is to switch places once the pulse starts its journey.
Setup-II
In this setup instead of measuring speed with delta-position and delta-time, use a property of pulse which indirectly depends on its speed (momentum with presumed directions, energy with full absorption etc.). So arrange interactions in three places: Here and now; a distant place where pulse moving away will reach and interact (then you will go there to read out interaction outcome), here but at a later time after the outgoing is reflected back to here; here but at a later time after pulse is reflected three times etc. If all measurements are the same, then you are convinced that the speed of light is what I measure here or there and independent of pulse being away or towards.
Note: It seems to me that it is almost impossible, in certain situations at least, to conduct a measurement of absolute significance that is INDEPENDENT of your theory or implicit assumptions. Roughly, I am telling that sometimes part of reality that we try to independently verify would actually be unreachable and we need a priory assumptions about it. 
