Cities are viewed as lively systems with internal flow structure of people, energy and goods. By appropriately defining concepts as metabolic rate, pulsating period, and mass previous studies have shown that cities follow known allometric laws of animals, then allowing rules for self-organization. The distributions of street lengths and nodes follow inverse-power distribution laws. That means that the smaller the network components, the more numerous they have to be. In addition, street networks show geometric self-similarities over a range of scales. Based on these features many authors claim that street networks are fractal in nature. What we show here is that both the scaling laws and self-similarity spring out of the underlying dynamics, together with the purpose of optimizing flows of people and goods in time, as predicted by the Constructal Law. The results seem to corroborate the prediction that cities fractal dimension approaches 2 as they develop and become more complex. Examples from History show that the Constructal Law provides a basis for understanding the internal structure both of roman and medieval cities. Finally, it is shown that the Constructal Law also provides the base for the design of flow structures (street networks, corridor networks) which perform optimally with respect to evacuating people from disaster areas.