We introduce the Xdπ calculus, a peer-to-peer model for reasoning about dynamic web data. Web data is not just stored statically. Rather it is referenced indirectly, for example using hyperlinks, service calls, or scripts for dynamically accessing data, which require the complex coordination of data and processes between sites. The Xdπ calculus models this coordination, by integrating the XML data structure with process orchestration techniques associated with the distributed pi-calculus. We study behavioural equivalences for Xdπ, to analyze the various possible patterns of data and process interaction.