I have problems in understanding why the following mapping from an ER-diagram to an ERM is correct, or, to be more precise complete. In the given example we have a ternary 1:1:N relationship between a project, a place and persons.
Each entity has a primary key (ProjectID, PlaceID, PersonID). If I understand this diagram correct the combination of a person and a project can not get associated to more than one place. And in addition the combination of a person and a place can only be associated with one project. Furthermore a project at a specific place can have multiple persons.
This understanding of how to read a ternary relationship leads to my problem. I map the ERM to the following RM:
Project(ProjectID) Place(PlaceID) Person(PersonID) Works(ProjectID, PersonID, PlaceID)
I now have two candidate keys in the table Works: (Place, PersonID) and (ProjectID, PersonID). Let's choose the first one as primary key. I should then have a correct RM (literature told me) but what I don't get is how to ensure that the same combination of person and project is not associated with different places? Don't I have to say somewhere that (ProjectID, PersonID) is also a candidate key or is this not part of the RM notation?
ProjectID PersonID PlaceID 1 Marvin New York 1 Tom Paris 1 Marvin Tokyo