Algorithms are developed for broadcast in hexagonally connected point-to-point networks. The algorithms are based on a modification of virtual cut-through switching. The authors advertise that this class of algorithms delivers multiple copies to every target node through disjoint paths. If they are right in this claim, then these algorithms are highly reliable in the presence of random unknown faults.

The authors attempt to verify the correctness of their algorithms by simulation in small hex-connected networks. They assume that the network is uniformly loaded, so the probability that a message is buffered instead of cut-through is the same at each node. It would follow that buffering frequency would be binomially distributed, but this assumption is unlikely to be valid in large networks. Further, a Poisson arrival process is assumed at source nodes (also unlikely in a real network).

The authors’ analysis of fault cases is simplistic, and I do not believe they establish that their algorithms are reliable in stressed operational environments. The paper does not convince me that the author’s algorithms actually deliver even a single broadcast copy to every reachable node, much less multiple copies.