For feature identity management we need something more precise than the rather loose semantics and grammar, defined only in natural language descriptions, which are commonly used in metadata descriptions. If a dataset is composed of several persistent feature collections then each could contain its own metadata, and in the limit of granularity, every individual feature could contain metadata on how it was constructed and under what conditions its attribute values were originally measured.
There are geographic-specific metadata protocols and systems
[7] as well as names and types [6], but the
recent enormous growth in non-geographic metadata protocols such as RDF
[8] implies that the GIS-specific protocols will have a
short life before they are merged into the mainstream.