Wireless sensornets provide an unprecedented opportunity to gather huge volumes of data about the physical world around us. Their severe resource constraints, however, in terms of power, memory, and bandwidth, have led to tightly-integrated and monolithic system stacks that made them difficult to program and limited code reuse and interoperability. As such, Culler et. al proposed an overall sensornet architecture that focused on a modular decomposition. In this work, we evaluate our progress in the two years since outlining our original vision. We use our two main architectural components as case studies, examining our successes and failure, and as such propose a new set of design principles that consequently alter our vision for the future of our architecture.