Abstract

Yahoo! Pipes is a well-known, widely used visual programming environment for creating data mashups by aggregating, manipulating, and publishing web feeds. It provides a natural laboratory for observing a range of end-user programming (EUP) behaviors on a large scale. We have examined more than 30,000 Pipes compositions in a search for regularities that might inform the design of EUP systems and their services. Although Pipes primitives span a broad range of functionality and can be richly parameterized and composed, we find a number of patterns that govern the structure and parameterization of Pipes in the wild. Most users sample only a tiny fraction of the available design space, and simple models describe their composition behaviors. Our findings are consistent with the idea that users attempt to minimize the degrees of freedom associated with a composition as it is built and used. 1.

Citations