Although sensor network research has grown tremendously in the past few years, it grows in the shadow of several long standing research communities, such as operating systems, networking, and computer architecture. Although few would argue that sensor networks are not qualitatively different than Internet-based systems, the question remains of whether they truly require new solutions, or can merely re-use the decades of work that precedes them. Starting from an architectural perspective, I'll show that, although the two classes of networks are different, many of their basic design goals are the same or similar. Correspondingly, many problems in sensor networks have Internet analogues, which are useful to borrow from but impossible to apply directly. I'll discuss what common (but not omnipresent) assumptions in Internet research -- beyond simple and well known ones like throughput and cost metrics -- generally emerge as stumbling blocks to producing sensor network solutions. Finally, I'll present a few examples of what I think are obscure (and possibly incorrect) parallels between active research areas in the two fields.