The Zipf's Law is usually invoked to describe those phenomena where it is natural to make an ordering procedure with respect to the ranking, for instance with respect to the size of cities, the amount of money of salaries, the length of rivers, etc. According to most of the scientific literature of this field the Zipf Law's is somehow considered "universal", in the sense that the details of the dataset are not crucial in order to obtain a Zipf's Law behavior. We will point out some simple considerations which suggest instead that details matter to obtain a power law in the rank-size plot and weaken the idea of the Zipf's Law behavior as a universal (natural) law for those phenomena where we can classify the elements of a set according to their size rank. Furthermore the emergence of Zipfian rank-size law in phenomena such as city size, salary, etc is still unclear and this field lacks of convincing models. We will introduce a probabilistic model which reproduce the empirical rank-size rule of city size and will outline a microscopic explanation to interpret the origin of the underlying distribution for city size that emerges in the framework of this phenomenological probabilistic model.