k-Set Agreement in Communication Networks with Omission Faults

We consider an arbitrary communication network G where at most f messages can be lost at each round, and consider the classical k-set agreement problem in this setting. We characterize exactly for which f the k-set agreement problem can be solved on G. The case with k = 1, that is the Consensus problem, has first been introduced by Santoro and Widmayer in 1989, the characterization is already known from [Coulouma/Godard/Peters, TCS, 2015]. As a first contribution, we present a detailed and complete characterization for the 2-set problem. The proof of the impossibility result uses topological methods. We introduce a new subdivision approach for these topological methods that is of independent interest. In the second part, we show how to extend to the general case with k in N. This characterization is the first complete characterization for this kind of synchronous message passing model, a model that is a subclass of the family of oblivious message adversaries