Georgi and Tom,
Thinking of Rob McEachern's argument that what we measure is simply that which we seek, the question of 'what is a photon' is a matter of the experimenter's choice. While I can go along with there being non-detected properties, and a divisible quantum or specifically, Quanta, at some point there is a need to mathematically qualify what it is that we seek.
In my early modeling, I simply followed a naive mathematical rationale which partitioned the individual Planck value Quanta between a mass:energy particle and the impetus which accelerated it to a peak periodic velocity at mid-point of any wavelength, and let the chips fall where they would. That quickly led to a seeming contradictory condition wherein the particle portion exceeded the Planck value at twice the wavelength of the arbitrary benchmark wavelength. In the same wise of a 'fact finding' exercise, I simply contented myself with an intuitive reasoning that the physical mass equivalence mattered less then the rate of momentum transfer by the particle form with any detection system. Crude and naive to be sure. But it, in the end analysis panned out as a parametric model leading to fully relativistic mathematical rationale which can account for the quantity of energy in a rest mass particulate field, distributed in a continuous gradient of density defining a finite volume. This satisfies Wheeler's zero boundary condition, where conventional attempts employing integrating over partial differentials always leads to a feedback loop at the minimum density boundary limit.
So I'm not the one whom should argue with how anyone else attempts to analysis "what is a photon?".
Best, as always. I've been dealing with some personal issues, so excuse me if I don't chime in very often :-) jrc