In this talk, based on a joint note with D. Buchholz, G. Ruzzi and E. Vasselli, it is presented an old standing question, initially proposed by J.E. Roberts in the '70s, about the nature of the operator associated to linked loops in QED. On the base of an homological argument, it is proved that some commutators associated to this loops are central in the algebras of the electromagnetic potential. Some consequences about the general theory of QED are also outlined.