Dataset

A ``dataset'' has an obvious meaning when a simple GIS organises its persistent storage as files. However, large and complex GIS applications involve databases and possibly a large number of files for import, distributed update, etc. Thus we need a tighter definition or a different concept entirely.

The OGC uses the term ``Feature Collection'' to mean an object which represents a collection of features but which also may support its own attributes. The fundamental operation on a feature collection is to make available (in some way) the features in the collection. Feature collections many be permanent, e.g. a dataset, or transient, e.g. the result of a query. The metadata of a dataset thus become attributes of a feature collection. Compatibility with other standards for metadata content can then cause problems because some which come from bibliographic communities allow ``repeated fields'' with different data, whereas most GIS data models for features (and an OGC feature collection is a variety of feature) allows only ``name = value'' semantics.

The existing OGC ``Simple Features'' specifications do not need to go into details such as defining the feature type of a feature collection [4] but the OGC Abstract Specification suggests that any feature could belong to several feature collections at once.