Thank you very much! If I well understood such constants are the best ones in the sense that for each bound it is possible to provide (independent and zero-mean) variables that saturate that bound, am I right? If so, is there any know families of r.v. where such best constant are known to be different from that of the larger class?