In this work, we advance the conceptual and technical aspects of Secure Multiparty Computation (SMC). We approach SMC as a computational problem and propose a novel formulation of this problem in terms of trust boundaries. From this formulation, we derive a general framework that enables a more comprehensive characterization of both the SMC problem and its solutions. Existing SMC solutions are commonly seen as diametrically different and incompatible, but we show how they can be mapped to particular instances of our framework, hence enabling their analysis under a common and unified basis. In this framework, the core component of an SMC solution is a distributed homomorphic cryptosystem. We show that the features this cryptosystem provides determine the need for interaction and overall efficiency of the corresponding SMC solutions. Based on this analysis, we introduce a practical instantiation of our framework by proposing a distributed version of the Brakerski-Fan-Vercauteren (BFV) lattice-based homomorphic cryptosystem. We analyze the security, noise overhead, and computational costs of this scheme. Due to its conceptual simplicity and efficiency, our solution has great potential for addressing highly relevant scenarios, such as secure data-sharing and machine-learning. Hence, this work constitutes a step forward in secure computation, by enabling computation across trust boundaries.