What chinglu hasn't realised is that you can give light coordinates only if you give a light-propagation event some coordinates, which is just what Einstein does in his paper.
Include the fact that light has a constant speed for all observers in relative motion, and that you have a reliable clock, you can calculate where the light is at some time after the initial event. Likewise, you can calculate where light has come from, but you have to detect it, put another way, you can't see light moving, only "leaving or arriving".
So if you know you emitted some light at (t,x,y,z), you can calculate the size of the expanding wavefront at t' and the distance it is from you; if you know there is another observer and where they are, you also know when they will see this light according to your own clock.
The only way to give light coordinates in any sense of the word is to generate some or 'absorb' some, and then these coordinates are only fixed by your local position. In Minkowski space 'distance' is not Euclidean; one needs to understand why light has no proper time in this geometry.
Of course, light moves through space "in" time in the Euclidean sense where space and time are both positive, but in Minkowski spacetime, you subtract the positive spatial components from the time component (or vice-versa), and the geometry is "decided by" the constancy of lightspeed which translates to a line with a slope of 1 (or 45[sup]o[/sup] from horizontal); so it becomes "one light second per second".
You don't have to generate a massive particle to give it some coordinates, now do you?