This is only my opinion of what a photon is...
If we consider a multi-dimensional hyperspace (more than 4 dims, but total quantity of open & closed dimensions are not pert. to the picture) - the photon, any photon, is a closed reaction that spans all of the volume of space-time (including time). The bulk majority of the history of each photon is along a small 4-volume of space-time between the reactions start and endpoints. The potential of the existence (the strength of effect of the electromagentic force associated with it) drops off very quickly away from this volume... by the mathematical relation described by the bell curve for the typical Gaussian wave-packet--- at a given cross-section through the four-volume between interaction points. Over large "stellar" distances, the packet becomes smoothly dispersed.
The reaction doesn't dominate the space-time, obviously, because it is of a very discrete property carried along the histories in the probablitity of the amplitude defined by the packet. I don't personally believe, however, that is is only probability distribution... Like Schroedinger, I cling to the belief that it is a real but behaves like a probability. People like Wheeler hate that kind of thinking. But if the EM force carried is some kind of kink in space-time dimensions stretched out over large distances... It would still obey statistics if the reaction had to be complete only at the endpoints.