DHS data chief says DRM needs more refinement
- By Joab Jackson
- Nov 04, 2004
The Office of Management and Budget's just-released Data Reference Model needs some adjustments if it is to help agencies share information, the metadata guru for the Homeland Security Department says.
'The DRM is a very good document,' Michael Daconta, metadata program manager for DHS, said at the National Science Foundation's recent Collaborative Expedition Workshop in Arlington, Va. But it needs further clarification, he said, because the model 'will cause confusion in its current form.'
The problem with the model, Daconta said, is that it doesn't follow the 'modern principles of knowledge representation.'
The basic DRM defines data according to three sets of characteristics: the context of the data entities being described, the characteristic elements of those data entities and the methods available for sharing that data.
Daconta suggested use of additional subcategories for each of these areas, changes he said would not affect the basic architecture of the DRM itself.
For instance, he said, the DRM needs a way to characterize associations. An association defines the relationships among data entities. Although the DRM suggests that associations can be described as a characteristic of the data, Daconta argues that associations should be separate. Associations really exist between different data entities, rather than as a characteristic of any one data entity, he said.
Daconta is working on a standard set of associations for DHS. Establishing formal associations will be essential for more intelligent sharing of data, he said.
'How many times have we said 'Why can't we just connect the dots?' We will never be able to connect the dots until we get serious about standardizing associations,' Daconta said.
The consistency of the DRM's language also needs work, he said. Some terms are not as rigidly defined as they could be. For instance, 'type' and 'object' are used interchangeably. Yet, in the programming world, those terms mean entirely different things, Daconta said. Also, there is inadequate language for describing an 'instance,' or a specific iteration of a type.
'It's important to be consistent and not use different labels when you mean the same concept,' he said.
Others at the workshop had some reservations about some of Daconta's suggestions. The Interior Department has implemented the DRM for several projects, and the lesson the architects of those projects have learned is to keep the model as simple as possible.
'When you try to sell a broad DRM to individuals who are not technical, it needs to be as simple as possible,' said Craig Tanner, a contractor who works with Suzanne Acar, Interior's data architect. Tanner works for IT Pioneers LLC of Charlottesville Va. 'The original DRM was simple enough, had a simple structure. It allowed us to implement this.'
'We don't disagree with Daconta,' Tanner said. 'We need to make sure we maintain simplicity.'
Joab Jackson is the senior technology editor for Government Computer News.