Geospatial Region Bounding Box skos:definition is wrong #913
Replies: 4 comments 2 replies
|
Regarding the bug, I suggest the immediate change to the definition to be:
I will make this change and push. However, I will also move this to a discussion since there are several additional issue that need addressing. First, the class "Geospatial Polygon" is incorrectly named. Polygons are 2-D planes. This definition seems to be referencing what is called a "closed polygonal chain", where a polygonal chain consists of connected line segments and closure indicates when the first vertex of such a chain intersects with the last. Second, reference to "polygon" runs the risk of confusing information with that it's about. Polygons are mathematical representations; they are not identical with the immaterial entities polygons are putatively about. We should avoid using information language when describing fiat boundaries. Similar remarks apply to "bounding box" is a geometric term. Third, 0, 1, and 2 dimensional fiat boundaries are boundaries of higher dimensional entities. For example, two 0 dimensional boundaries are what bound a 1 dimensional boundary, 1-D boundaries bound 2-D boundaries, 2-D boundaries bound three dimensional volumes, i.e. sites. Strictly speaking, lines bound surfaces, not sites like Geospatial Region. My suggested revisions to kickoff a larger discussion:
@alanruttenberg @APCox @michaelrabenberg @CarterBeauBenson @cameronmore @mark-jensen @neilotte |
|
Do we need a "boundary of" object property? Otherwise, I suggest using continuant part of instead. Additionally, is a fiat line that extends horizontally across the full length of a fiat surface a vertical boundary of that surface? |
|
On reflection, I'm struggling to see what the "that has another Geospatial Region as a non-tangential proper part" part of the GRBBox definition adds. Pick a site, any site, it'll have a non-tangential proper part, won't it? So if that clause were struck from the definition, I at least don't see how that'd make any extensional difference. So why keep it? Perhaps one might say: No, sites don't always have non-tangential proper parts; you can't just gerrymander proper parts of sites in the way that'd be required for that to turn out true. But if that's right, then presumably there's no guarantee that, for any Geospatial Region, r, that some GRBBox is a BBox "of," there's a Geospatial Region of which r is a non-tangential proper part. But the definition clearly presupposes that, for any Geospatial Region, r, that some GRBB is a BBox "of," there's a Geospatial Region of which r is a non-tangential proper part. So denying that every site has a non-tangential proper part would generate problems of its own for the definition. Perhaps the intuition is supposed to be that a GRBBox is a BBox "of" something strictly within it, so you need something about "non-tangential" in the definition. But two things. First, are BBoxes typically understood not to touch what they're BBoxes of? A little googling made me doubt that, but that's the extent of my acquaintance w/ the subject, so take that for the very little it's worth. Second, even if that's right, I think that'd be a reason to have something about "non-tangential" in a bounding-box-of relation; I don't think that'd be a reason to have such language in the definition of GRBBox, given the above considerations. |
|
The definition of Geospatial Region Bounding Box, and the proposed updates are inconsistent with the use of 'Bounding Box' in the geospatial community. The term 'Bounding Box' is generally used synonymously with Minimum bounding box. It defines the minimum and maximum values of the set of coordinates in a geospatial object. It is not an arbitrarily shaped boundary, and it is not a line. It will have the same dimensionality as the object. It can be visualised as a (hyper)rectangle - i.e. a 'box'. In the tabulation above, and in the CCO GeospatialOntology module, the class name Geospatial Region Bounding Box is thus likely to be misleading to typical geospatial practitioners. (I thought I had submitted a brief version of this comment as a response to the email notification, but it doesn't appear to have stuck.) |
|
In other words, the class definition is fine. But if it is a generic closed fiat line, then this is not what is commonly understood to have the name 'Bounding Box'. |
|
As explained above, the new definition of Geospatial Region Bounding Box is inconsistent with the common usage. In the geospatial community, a bounding box can be viewed as a rectangle (or 1-D or 3-D version of a rectangle) whose sides or faces are aligned to the coordinate axes. The key capability is that it provides the minimum and maximum values of the coordinates of the thing, which can be used for indexing and discovery. Strictly it is not actually a rectangle, it is just the limiting coordinate values in the thing bounded. But if visualized as a rectangle: yes, it touches the thing it bounds. It is not a polygon. The previous definition at least allowed for the conventional understanding. |
Uh oh!
There was an error while loading. Please reload this page.
Here's the skos:definition:
A Geospatial Polygon that has some Geospatial Region as a non-tangential proper part.
cco:GeospatialPolygon is a subclass of obo:fiatLine. cco:GeospatialRegion is a subclass of obo:site. If x has nontangential part y, then x has continuant part y. (hasNontangentialPart is a subproperty of hasSpatialPart, which is a subproperty of hasContinuantPart.) So the skos:definition implies that a Geospatial Region Bounding Box is a fiat line that has continuant part some site. This can't be. Fiat lines are 1D and sites are 3D, so the dimensionality facts alone preclude this; and at any rate [cwp-1] from clif bfo prohibits a fiat line from having a site for a continuant part.
I don't know what the best replacement of this definition is. I assume something mushy like "that bounds some Geospatial Region" would at least be an improvement.
All reactions