gistfile1.txt

Some type of (graph) database which I wish existed.

Assume I've put Wikipedia into the graph database. I'd like to ask questions
like "What are American movie stars under 39 years old with over $1B in movie
grosses?". It would be acceptable to make this a prepared statement of the
form "?nationality movie stars under ?age with over ?dollars in movie
grosses". I believe this is like SPARQL?!

Nodes in the graph can be of many "classes." A class would be a bag of
attributes that the node should respond to, similar to Ruby modules, or Java
interfaces. I would be a Person, Programmer, Male, etc.

At insert/update time, we can maintain indexes designed to match the prepared
statements. Each prepared statement could register an update handler (in this
case to the "movie star" and "movie" classes). When I insert another movie,
the index is updated. Ideally, the index could be distributed a la Cassandra's
OrderedPartitioner, so that I could both scan and lookup over the index.

Check out Blueprints [http://blueprints.tinkerpop.com] and then bindings to Neo4j and OrientDB.