We recently introduced a new ensemble of low-density parity check codes
with the following properties: 
1) It achieves capacity on the erasure channel, although thanks to a
   quite different mechanism with respect to LT and Tornado codes.
2) It achieves capacity on a probabilistic model for a rank metric 
   channel. The latter was advocated by Koetter and Kschischang as
   a model for errors in network coding.
In this paper we investigate several properties of this ensemble with 
a special focus on complexity and implementation issues.