In a recent post Freek Lijten shares a simplified version of a solution he came up with to separate models and their logic between the fetch and save operations.

Basically I am wondering how storing data and retrieving data are different and how you should / could model this. Since we need different information while storing and retrieving information it makes sense to model those actions in different ways. The how is still bothering me and in this post I give a possible solution. I truly hope I can get some reactions and thoughts of other people in on how they would solve these kind of problems.

He introduces the basic structure of the application he's working with and how the concept of "documents" ties in. Then he gets into the problem: the differences in data required for the save versus locate and load. His solution is to split out the different pieces (relations) of the document into separate value objects. These objects then only contain the handling to get only the relations needed on the load. He doesn't like the solution, however, because of the amount of overhead it introduces.