The persistent object store whose design criteria and initial prototype are described here is part of a larger project to develop techniques for integrating programming languages and database features. Moss claims that separation of the programming language from the database function invariably leads to poor performance.
In order to build the prototype, a number of decisions about design had to be made, such as where the detailed object semantics would be implemented and why. The designers chose to implement it in the user address space, making changes in policy easier. The paper contains a good discussion of this and other design choices, and it is not difficult to read.