Version 5 (modified by jbourne, 16 years ago)

--

POSC-Caesar FIATECH IDS-ADI Projects

Intelligent Data Sets Accelerating Deployment of ISO15926

Realizing Open Information Interoperability


RDS/WIP World View: Models, Meta-Data and Data

The convention when modeling information is to divide it into model and data, where the model is the schema that governs the form of the data and against which it can be validated.

Early Binding

The temptation when implementing a system is to embed all knowledge of the model in the system, but to leave the data free. The problem with this approach is that it does not allow transmission of models, only transmission of data. We term these kinds of system "early-binding" which means that they deal with a fixed model.

Late Binding

In contrast, in a collaborative reference data library, no system can truly embed all knowledge of the model, because the model is always expanding. This means that sometimes, model needs to be transmitted as data. We term these kinds of system "late-binding", which means broadly that they can deal with ad-hoc structures.

Meta-data is Data

To take this a step further, in order to validate data against a model in a late-binding system, all of the meta-data about the described structures also needs to be transmitted as data.

These leads us to the conclusion that meta-data and models are also just data, and since a machine needs to embed knowledge about the structure and meaning of data in order to process it, we are therefore well situated by representing all data in the same way - models and meta-data look just like data to the machine.

The Role of RDF

For the RDS/WIP we have chosen RDF as the common representation because it reduces all information to simple statements of exactly the same form.

In concert with SPARQL, this provides a means of being able to query information about any model (or data in any model) without the need to know anything about the structure of that specific model beforehand.

The Limitations of RDF

RDF is however a highly atomized view of data and so both projections of other forms to RDF and from RDF must also be supported by the system to properly address the problem.


Home
About PCA
Reference Data Services
Projects
Workgroups