graph databases

Scott Hanselman speaks to Haixun Wang on Trinity (which is to be Microsoft's Graph Database) in this podcast. We are likely to use a graph database on our project. The most important thing to know, per Mr. Wang, is that the graph database does not use joins which are "expensive." From a point in a graph database, one may find a neighbor without a key join.

We are to have a fascinating dynamicesque feature in our app which will allow users to bolt onto objects custom attributes. While the base objects will be kept in a SQL database, the architects are looking for an alternative approach for the keeping of the custom attributes.

We are more likely to use neo4j for a graph approach. We may also not take a graph approach and take a Cassandra or Redis document database approach instead. The die is not yet cast.

Dunno.

Wikipedia sort of suggests that persisted items expose nodes which have "properties" that one may match against when reaching outward from other persisted items in searching to ultimately make matches to persisted items, grabbing hold by way of nodes (which are almost like the APIs persisted items expose).