I've been working on the next River product. This time I'm using a MySQL database. Three tables -- feeds, items and subscriptions. The folder structure is exactly as in River5, except there is no data folder (the data is in the database). I am still a relative newbie in SQL databases, but I think this model works. I'm documenting as much as I can and of course I will release the Node.js source. I hope it serves as a basis for distributing RSS intelligence around the net. Last time around (Google Reader) we centralized. That was a mistake. If enough people run instances of this database we'll have a less interruptable base of functionality. I want to try out more new ideas as well. We've been really stuck for a long time.