Is a specific cube-schema agnostic to what client will use it,; what queries will come? I mean is it reasonable to imagine a generic CubesNavigator, that let's a user look at any combo of properties and aggregations across any cube. That would require metadata queries like getDimensions(), getMeasures(), getHierarchies() to support populating dropdowns to start navigating down the tree. The reason I as; is because then Stefan, the back-end wouldn't know in before hand what you answered to a previous question: "Only joins necessary for your query are considered. It does not generate the whole star/snowflake for every query" Are you saying, you have to predefine what parts of the snowflake/star are available to query? Like "available slices"?