In the early days of computer-based map making, maps were made using a printing process. Each colour (black, blue, etc.) was separately applied as “layer” to the map sheets as they went through the printing press. The layer for each colour was defined by a “peel coat”, a plastic mask that excluded the areas which were not to be printed, and which was cut from a map file using a digital plotter. Layers were, quite literally, layers of colour, and people spoke of the “blue layer” (water), the “green layer” (forests, parks), and so on.
When digital mapping moved things to the next stage, the layer concept was retained, and early systems had a number of layers defined by the bit depth of the raster representation. Each layer retained a colour associated with it, just as in the map sheets from the old printing process.
Raster maps predominated in the early days of digital mapping. A raster map provided in effect the description of real world objects, but using what mathematicians would call a characteristic function (also known as an indicator function – see https://en.wikipedia.org/wiki/Indicator_function ), meaning that it was zero wherever the object was absent, and one wherever it was present. It made for fast drawing and overlay, but made it difficult to determine the length of a road or the shortest route between two cities. More explicit geometric representations, using control points with coordinates and interpolation, were introduced next and became known as vector encoding.
At this point, vector digital mapping and computer-aided design were very much the same; the only difference was in the kinds of coordinate systems being used, as digital maps tended to cover large areas of the earth relative to their CAD cousins. Neither environment had any notion of what we, today, would call a feature. Then a curious thing happened.
People began to use the term “feature” in reference to “things on maps”, such as a road feature, or a park feature. This did not mean a model of the object in question, but rather something that was distinguished or visible on the map. Similar terminology also appeared in the analysis of images, where again it referred to some visible artifact.
The term “feature” began to acquire the connotation of “object” in the early meetings of the OGC (93-94) and TC211, and was separately and officially formulated as such in the TC211 General Feature Model (ISO 19109), OGC GML 1.0, and in the OGC Abstract Specification. From that time forward, “feature” has come to mean a model of a real world object or phenomenon, which is characterized by a set of properties or characteristics. While it took a good deal of time for this notion to fully take hold (see the vestiges of features as map features in http://webhelp.esri.com/arcgisserver/9.3/java/index.htm#geodatabases/feature_class_basics.htm ), it is now generally accepted. Features are models of real world things or phenomena that may have zero or more geometry valued properties. In short, they are objects.
For some reason, this object-based approach took much longer to take hold in the domain of engineering and architectural drawing and design. While some early pioneers like Dassault spoke about Computer Aided Design (as opposed to drawing), and introduced early notions of objects, it was much later, with the rise of Building Information Models (BIM), that this began to take hold with respect to architectural and engineering structures. With BIM, we have much the same notion as feature with a building (or other structure) object being composed of other, more fine grained, objects such as walls, doors, and windows, and the BIM encoding expressing the spatial and other relationships between the component objects.
Most modern CAD and design tools now offer a BIM-based encoding, and concrete BIM encodings, such as OGC CityGML and Building Smart’s IFC, are increasingly wide spread. GML, in particular, offers a complete unification of the feature world (GIS, spatial DBMS) and BIM, since it supports the TC211 General Feature Model and also supports a complete range of geometric types from points, to geometric and topological solids. With GML, the feature notion has come into its full flowering and can be applied with equanimity to terrain, built structures, administrative regions, demographic data, images, and the description of the natural world.