Spectrum sharing is going to be essential for the efficient (and evolvable) use of radio spectrum among heterogenous uses. To support this, the regulatory system has to go beyond its current law-and-footnotes interface and include technical dimensions to its architecture. The key issue is "trust" and the overhead involved in maintaining it. Trust plays the role of the "thin waist" here, and our results suggest that by bringing in a mechanism-design philosophy, it should be possible to assure "trust first and performance later" in the same way that the secret to the Internet was arguably "connectivity first and performance later." Since this suggests a new potential audience for IT results, I will close with some natural information-theoretic open questions relevant to this problem.