Current optical transport networks provide high bandwidth through the use of advanced WDM technology, but are difficult to adapt to the different statistical patterns and quality of service (QoS) demands of future traffic. There has been much debate whether the use of dynamically reconfigurable optical networks would have a number of advantages in accommodating the needs of future traffic demands. Dynamic networks would eliminate the need for frequent opto-electronic conversion in current networks, and may save resources through higher utilization and fast adaptation. Different architectures have been proposed to address this problem: Wavelength-routed optical networks (WRON), optical burst switching (OBS), and optical packet-switching (with increasing granularity and speed of reconfiguration). In this paper we investigate whether these architectures are suitable (necessary?) to meet the demands of future traffic, using an analysis focusing on both modeling and experimental aspects.