These additional functions tend to involve several datasets with specific relationships, e.g. one dataset may be a prior version of another. We then require some way of tracking individual features across those datasets. The standard example is where some attributes of a feature are under the update authority of a different organisation from other attributes; but somehow all parties must agree that they are referring to the correct feature even if many (or all) the attribute values change.
A study of feature identity presupposes that we will be dealing with moderately persistent real world objects which are observable as distinct entitites (at least for a while): entities that exist long enough to be worth naming and talking about. Thus this paper is firmly placed in the ``object'' rather than the ``field'' tradition of GIS, with the proviso that some of these objects may have indistinct boundaries [1] and may be temporary, e.g. sandbanks, storms and forest fires.
This paper attempts to review conceptual structures which may underpin future interoperability standards. File data formats have a relatively short useful life compared to the life of the data they transport and ``standard'' function interfaces have even shorter lives, but the data model has a much longer life: almost as long as that of the data itself.