Currently when executing a query through the Magnolia query API, there is no way of preventing the result set from being iterated completely. When e.g. only the first couple of objects from a large result set are of interest, then thousands of objects are instantiated unnecessarily, resulting in a huge performance penalty.

These problems have been addressed in the openmind Magnolia Criteria API, but this cannot be used e.g. for the STK implementation.

So in order to optimize e.g. the STK, the core query API must be extended.