Just to make sure this is clearly understood: Two photons absolutely can collide head on, annihilating themselves to produce an electron-positron pair in their place. In principle, sufficiently high-energy colliding gammas could produce pretty much any particle pairs, but of course it's tough enough just to get two photons whose net energy is twice the mass of an electron.
Photon annihilation has to be true because of time symmetry. That is, since one of the decay modes for positron-electron annihilation is the production of two gamma photons, time symmetry requires that the inverse process also be possible. It's just harder to arrange, particularly since uncharged photons have no guiding attraction for each other.
This idea is not hypothetical. Photon annihilation was demonstrated in experiments done several years ago by shining a green laser directly into a beam of gamma rays. A few of the green and gamma photons mutually annihilated to produce fast-moving electron-positron pairs with leftover momentum in the gamma propagation direction. Looking for the reference now...
I think this was it, but it seems earlier than the one I was thinking of:
D.L Burke et al, Positron Production in Multiphoton Light-by-Light Scattering, SLAC, June 1997.
If you allow photon-photon interactions (scattering) via virtual particle pairs you get two-photon physics, which looks at the probabilities of photon-photon production of particle pairs much heavier than electrons.