Network coding for dynamically changing networks

We here advocate the case for network coding as a guiding paradigm for the operation of networks that vary in a small time frame, due to node mobility, channel variations, and varying traffic conditions. Three ideas that appeared succesively in time brought in place an elegant operation for network coding in such environments. These are, use of randomized network coding for intermediate node operation, use of generations to avoid synchronization, and use of subspace coding to allow for small packet sizes. Information theoretical performance limits support these results.