The growing market of wireless devices with multiple connectivity technologies, e.g., Wi-Fi, Bluetooth, and UMTS, is pushing toward the necessity of seamless autonomic selection of the proper connectivity network at any time. That selection should be context-dependent and consider several elements, at different abstraction layers, from bandwidth requirements to network congestion, from connectivity costs to user preferences. As a symptom of the recognized relevance of this topic, relevant research work has recently addressed context-dependent connectivity management. We claim the need to provide both practitioners and newcomers of this research field with a comprehensive survey with a unifying perspective, to better point out similarities and specific aspects of the numerous solutions in the literature. In particular, the paper aims to achieve the twofold goal of i) identifying the primary design choices, together with corresponding tradeoffs, emerging from recent proposals, and ii) better positioning the wide variety of related work based on a single classification scheme. To that purpose, we introduce an original model to represent solution architectures, by showing how it makes the adopted design choices manifest and facilitates the positioning of proposed approaches, e.g., by throwing new light on the usage of 4th Generation and Always Best Connected terms. In addition, based on that model, we propose a novel taxonomy to cluster the proposals in the field along three first-level classification directions: management scope, evaluation process, and continuity management. A wide set of examples illustrate how our taxonomy is useful to thoroughly understand the differences and similarities among state-of-the-art solutions in the literature, to some extent independently on technology-specific implementation details, thus providing a valuable classification tool also for future proposals.