This reminds me that you can do a lot without having to setup an RDBMS. Take Hacker News for instance. However, the OP uses a file as a table, whereas HN uses unix directories as tables and uses files as table records, afaik.
I tend to use sqlite for small projects, but I never really fancied SQL much and I'd never put an ORM on top of sqlite... so for really small things (and only local) I believe this may help
So what about managing indices (or at least map-reduce views - CouchDB style).
I'd really quite like the option of having something like CouchDB that kept documents as separate files but still allowed efficient querying.
Hey, that is reaally cool, despite I needed the native interface with python objects I think full bash sounds cool too ^^
Thanks! but I can only see it as a bad hack, as it's just grep doing almost all the dirty work :) btw I got the idea from http://zachholman.com and his boom-db...