I make the assumption that the analysis classes will not evolve into tthe implementation classes. That way I only add to the domain classes the attributes of interest from an analysis point of view. You should not be defining attribute types, etc. in the analysis phase unless these are in fact constraints on the design (e.g., Part Number values are represented as a formatted string values)
The other thing is during analysis the concept of an employee (has a position, a supervisor, a date of hire, etc.) might be understood by the buisness and represented as a single analysis class, while the implementation of an employee might actually be several classes (demographic data, employment/position data, reporting heirachy, etc.)
If later you find attributes from your domain classes, just select those attibutes from your domain class (project bowser side) and drag and drop them onto your implementation classes (and they will be added)
However, the naming conventions for my analysis classes are named to make since to the business users, while implementation classes are named following the project's coding conventions.
Just by 2/100's of a Euro,
David
