The purpose of data in a diagram-centric tool is to store enough information for the diagram to be redrawn and, possibly, to allow the user to add additional annotations. There are two serious disadvantages with this approach. First, if the same element being modelled appears in two different sorts of diagram the data about that element is likely to be disjointed and in two different places. Second, the data structure is likely to be optimized for diagram generation rather than for analysis or browsing.
Some existing standards are diagram-centric. An example is BPMN - Business Process Modeling Notation - from OMG. For instance, BPMN has a data structure called "Pool" used for modelling process diagram swim lanes. Swim lanes are usually used to indicate that a business process belongs to a particular department or organization, but the word Pool is clearly a diagram element not a business element.
The data-centric approach structures the data with the requirements of the information to the foremost not the requirements of the diagram. The advantages of the data-centric approach are:
The disadvantage is that it is more difficult to generate a diagram from the data, but that is the vendor's problem not the user's.