So I am mulling this real time streaming data concept when I come across this article from Forbes: Data On The Fly. It talks about Michael Stonebraker of Ingres and Postgres fame, who started a new company called Streambase. This outfit has produced a derivative of SQL they call StreamSQL, and the claim is it can process data as it comes in the pipe, before it is written to disk. Great for big hedge funds and market makers.
Seems like this StreamSQL could prove a pretty interesting tool for a peer-to-peer database network. Finding out what data you really need to hold on to before you write would fit into the limited storage capacity scenario likely existing at each node in such a network.