Network coding enables novel network functionalities and thus offers a wider canvas of choices when optimizing an information flow problem. In this paper, we examine the simplest possible information flow problem, a unicast connection, and explore what we believe is one of the most attractive features network coding offers: the ability to enable near-optimal performance in a completely decentralized and randomized setting. This is an especially attractive feature for wireless applications. However, it comes at the cost of an overhead in terms of rate that can be significant for applications that operate using relatively short frame lengths, as is the case in the wireless setting. We review the efforts in the literature to either alleviate this overhead, or alternatively, to exploit it for network management and control.