As freight deliveries in cities increase due to retail fragmentation and e-commerce, parking is becoming a more and more relevant part of transportation. In fact, many freight vehicles in cities spend more time parked than they are moving. Moreover, part of the public parking space is shared with passenger vehicles, especially cars. Both arrival processes and parking and delivery processes are stochastic in nature. In order to develop a framework for analysis, we propose a queueing model for an urban parking system consisting of delivery bays and general on-street parking spaces. Freight vehicles may park both in the dedicated bays and in general on-street parking, whereas passenger vehicles only make use of general on-street parking. Our model allows us to create parsimonious insights into the behavior of a delivery bay parking stretch as part of a limited length of curbside. We are able to find explicit expressions for the relevant performance measures, and formally prove a number of monotonicity results. We further conduct a series of numerical experiments to show more intricate properties that cannot be shown analytically. The model helps us shed light onto the effects of allocating scarce urban curb space to dedicated unloading bays at the expense of general on-street parking. In particular, we show that allocating more space to dedicated delivery bays can also make passenger cars better off.