We present a web service based approach to enable an evolutionary sensornet system where additional sensor nodes may be added after the initial deployment. The functionality and data provided by the new nodes is exposed in a structured manner, so that multiple applications may access them. The result is a highly inter-operable system where multiple applications can share a common evolving sensor substrate. A key challenge in using web services on resource constrained sensor nodes is the energy and bandwidth overhead of the structured data formats used in web services. Our work provides a detailed evaluation of the overheads and presents an implementation on a representative sensor platform with 48k of ROM, 10k of RAM and a 802.15.4 radio. We identify design choices that optimize the web service operation on resource constrained sensor nodes, including support for low latency messaging and sleep modes, quantifying trade-offs between the design generality and resource efficiency. We also prototyped an example application, for home energy management, demonstrating how evolutionary sensor networks can be supported with our approach.