Developing algorithms for sensor networks—and proving their correctness and performance—, requires simplifying but still realistic models. This paper surveys various models in use today and puts them into perspective. In addition, we propose interesting models which are not widely adopted by the community so far.