In this paper we consider the design of linear multicast network codes, and in particular examine the tradeoff between knowledge of network topology and performance of the resulting network codes. We focus on the distributed design of zero-error multicast network codes. We show a simple scheme when the multicast capacity equals $2$. We then consider general multicast problems and give two algorithms for such problems. Our algorithms require only that each node in the network possesses an identification number that is unique to it. Our algorithms and bounds presented below are also of independent interest for the field of theoretical computer science.