Stone is a cool idea. I like the no database idea. I remember being at RailsConf 2007 and hearing Avi Bryant talk about SmallTalk and his database less SmallTalk web framework called seaside, http://www.seaside.st/.
Later Dave Thomas gave his keynote and asked a similar question of are relational databases the way to go? Why aren't we trying out OO DBs? It would be nice to see some benchmarks on Stone, anyone know what speed for it is like?
I quickly read over the source code, and it looks like the database is backed by a directory of YAML files. That basically means you get O(log n) performance for random lookups (based on the underlying filesystem, of course), but O(n) searches.
I use this approach for Angerwhale (my blog software), and it mostly works fine because O(100) is always going to be fast :) The reason I didn't use an RDBMS is beacuse I wanted all the data to be editable in a text editor (so I wouldn't have to write management tools). I think I might add optional indexing at some point, though, just to make it a bit more scalable.
Later Dave Thomas gave his keynote and asked a similar question of are relational databases the way to go? Why aren't we trying out OO DBs? It would be nice to see some benchmarks on Stone, anyone know what speed for it is like?