It now seems clear that the Smart Dust vision of the late 1990s was well ahead of its time. It may also be argued that
the challenges of extreme scalability were vastly underestimated. The transition from hundreds to thousands of nodes was
achieved with multi-tiered networks within the scope of early efforts, such as the DARPA-funded ExScal project [123] of a
decade ago, but was virtually ignored in subsequent work. Similarly to scalability, miniaturization was demonstrated over
a decade ago with the Spec mote design [12] and later ignored to concentrate on matchbox-sized platforms [21]. The same
goes for application scenarios: sensor nodes were dropped from aircrafts at the Twenty nine Palms DARPA demos [124]
over ten years ago to showcase a scenario recurrently mentioned in the literature at the time [29], but today’s tendency is
to build sensors into unmanned aerial vehicles rather than to drop them. While failing to become truly pervasive, however,
wireless sensor networks have demonstrated their role as a powerful tool for domain scientists whose applications do not
require or do not want humans to be actively in the loop. Going forward, we expect them to retain this specialized role.