How to deliver data to, or collect data from the hundreds of thousands of sensors and actuators integrated in "things" spread across virtually every smart city streets (garbage cans, storm drains, advertising panels, etc.)? The answer to the question is neither straightforward nor unique, given the scale of the issue, the lack of a single administrative entity for such tiny devices (arguably run by a multiplicity of distinct and independent service providers), and the cost and power concerns that their direct connectivity to the cellular network might pose. This paper posits that one possible alternative consists in connecting such devices to their data collection gateways using "oblivious data mules", namely transport fleets such as taxi cabs which (unlike most data mules considered in past work) have no relation whatsoever with the smart city service providers, nor are required to follow any pre-established or optimized path, nor are willing to share their LTE connectivity. We experimentally evaluate data collection and delivery performance using real world traces gathered over a six month period in the city of Rome. Results suggest that even relatively small fleets, such as an average of about 120 vehicles, operating in parallel in a very large and irregular city such as Rome, can achieve an 80% coverage of the downtown area in less than 24 h. © 2016 Elsevier B.V. All rights reserved.