Discussion to: "Old and recent results on the relationship between predictive inference and statistical modelling either in nonparametric or parametric form" by E. Regazzini