Researchers report on their success in applying two "hyper" topologies to improve the scalability of Gnutella. This brief paper seems to be getting a lot of press.

Doesn't this seem like a "duh" to you? These topologies aren't new. I recall discussing the virtual hypertorus with Tom in ~1996 in regard to the "pipenet" idea. They are quite common in the literature of high-performance cluster computing.

In an ad-hoc P2P network, I suspect it will be problematic to discover the network topology and then force users to interconnect themselves in this way. Even so, this paper is useful in that it demonstrates (theoretically) viable alternatives to the untrustworthy Morpheus/KaZaA-style "supernode"-based architectures.