This paper presents IP(dmux) - a network-layer infrastructure that serves as a general-purpose deployment vehicle for next-generation network protocols. For each new network protocol, IP(dmux) provides a network-level access path between any end-user and the (approximately) closest router supporting the new protocol. IP(dmux) thus ensures that even partial deployments of new protocols are easily accessible by the global Internet user population. We present the design and implemention of IP(dmux) which we then use to experiment with three next-generation IP architectures - IPv6, FRM (a new protocol for global network-layer multicast) and i3 (a rendezvous-based network architecture). Our experiences suggest new networklayer architectures can easily leverage IP(dmux) to aid their deployment. Moreover, our evaluation through simulation, measurement and wide-area deployment indicates that even small-sized IP(dmux) deployments can yield reasonable endto- end performance for partially deployed next generation architectures.