We discuss some of the preliminary results and findings derived from the analysis of a first sample of flares detected by the XRT on board Swift. The analysis shows that the morphology of flares is the one we expect from the collision of ultra-relativistic shells as it happens during the internal shock model proposed by Rees and Meszaros in 1994. Furthermore the Delta(t)/t ratio and the decay-time to rise-time ratio have mean values that are in good agreement with the values observed in the prompt emission pulses that are believed to originate from internal shocks. The conclusion is that the flare analysis favors the internal shock as due to shells that have been ejected by the central engine after the prompt emission. The central engine seems to remain active and capable of generating large amounts of energy also at later times. More data are needed to establish whether or not some of the flares could be due to shells that have been emitted with small Lorentz factor at the time of the prompt emission and generate later time flares due to the catch up of these shells.