Interesting idea! I wonder how they think to address performance (one data query needs many requests to different nodes if I understand it well) and availability (if one node is down, I can't access my data?).