Street networks are one of the very few types of complex networks where the history of the network can be traced over long periods of time. Here we introduce methods for quantifying the geometric characteristics of street networks and analyze the details of the evolution of the networks of Sheffield (UK), Khorramabad (Iran), and Kerman (Iran) over tens to hundreds of years. The results suggest that the spatial distribution of streets, in particular their lengths and orientations and associated entropies, are strongly affected by the surrounding landscapes but in different ways. The length entropies are mostly controlled by the space available for network growth, whereas the orientation entropies are mostly controlled by the shapes of the constraining landscape features. Thus, the network of Khorramabad, located in a narrow crescent-shaped valley, shows the smallest increase in length entropy (space constraints) and the largest in orientation entropy (shape constraints) over time. By contrast, the network of Kerman, in a flat desert area with essentially no space or shape constraints, has the largest increase in length entropy and the smallest increase in orientation entropy over time. The network of Sheffield, a city subject to moderate landscape constraints of hills and valleys, has length and orientation entropies that fall between these extremes.