Gavin Salam's talk at the "Altarelli Memorial" session of the ICNFP 2016 conference, which is presently taking place in Kolimbari (Crete), was very interesting and I wish to report here about it. 
As this was a session focused on remembering Guido Altarelli and his contributions to theoretical high-energy physics, Salam started his talk by showing in quantitative terms just how big was the impact of Altarelli on the research that is done nowadays at the LHC. He produced plots of the most cited works in the >1000 publications by ATLAS and CMS, showing that the physics we do there is deeply founded on the knowledge of the structure of the proton, to which Altarelli gave foundational contributions.

The second part of Salam's talk, however, contained surprising new developments in our understanding of "the last parton" in the proton - not a quark or a gluon, but the photon, the quantum of light. This work was produced by Salam in collaboration with Aneesh Manohar, Paolo Nason and Giulia Zanderighi. 
Of course, the proton contains photons, as in its interior there are moving electric charges (the valence and sea quarks), and the acceleration of electric charges always involves photon emission. 
Just how probable it is to interact with a photon when you probe the proton's interior, however, is not very easy to determine. Interactions with a proton mediated by photon exchange may be elastic, when the proton bounces off, or the breaking up of the proton; these are called inelastic interactions. It is important to derive the probability that these processes occur, as a function of the energy of the interaction.

The situation can be pictured as shown below, where the uncertainty in model-independent determination of the parton content of the proton is plotted as a function of x, the parton momentum fraction, for the well-known up quark, the so-so-known strange quark, and for the photon. As you see, the photon content of the proton is really not very well known!

Salam showed how to make progress. One can study a hypothetical process whereby a neutral heavy lepton is produced by lepton-proton collisions. The cross section for this process can be written exactly as a function of F2 and FL structure functions describing the hadronic tensor of the proton source. But one can equally write the process in terms of the interaction of a photon source within the proton with the lepton. 

Above, one of the processes considered by Salam in the calculation of the photon PDF. A heavy neutral lepton (thick red line) is produced by the interaction of a lepton (narrow red line incoming from the left) with a proton-emitted photon (the wiggly line at the center).

The matter is too advanced to be described here, and I myself have trouble understanding the details - imagine explaining them to laypersons. So rather than going into the details, I only mention that this "theoretical trick" allows one to relate the photon PDF to the FL and F2 parton distribution functions. The result of the calculation - the LUXqed predictions - is an amazing precision in the photon PDF, as shown in the graph below.

It was very nice to hear Salam's description of this new development in the knowledge of the proton structure during a session commemorating the legacy of Guido Altarelli. So now we know with very good accuracy how much light do protons contain! If you think about it, it's awesome: we are made of protons, and protons are, in some part, made of light... And now we know how much of it.
UPDATE: The article describing the research summarized above has been published on the arxiv on 7/16 and is available here.