Storing info from DOM in a Content Repository

I'm investigating the pros and cons of storing information from a DOM as
well as associated metadata in a content repository, specifically a JSR 170
implementation.  The one I'm looking initially is Apache Jackrabbit, but may
be looking at Magnolia and others as well. Basically I want to see how this
stacks up against (a) storing same info in a relational database, and (b)
storing same info in a native XML database. Anyone know of similar work
already performed, or have any comments/suggestions?   

Thanks,

Bill Barnhill