Conventional data networks are based on layered architectures. The introduction of wireless networks has created a need to violate this layering discipline to create cross-layer designs or adaptations. Ad-hoc implementations of such cross-layer adaptations reduce the level of modularity and abstraction in the network?s implementation, giving rise to a significant increase in complexity. We present a taxonomy of possible cross-layer adaptations that is then used to derive an architecture for their implementation that significantly preserves the networks structure. We present implementation results that validate this architecture in the context of a real wireless network implementation.