WG9: Replace ISO 19125-1 with multi-part standard about information management

[gbuehler]

Justification: ISO 19125-1 needs to be revised following revision of ISO 19107. Proposal: Replacement of ISO 19125-1 with multi-part standard about information management.

[gbuehler]

To be discussed at OGC meeting in Wellington (Dec 2017). Replacements to be consistent with SQL-MM. New name and number to be decided. From the Wellington JAG report (N4706):

• To be developed in a multi part series of standards
• New approach based on features and their description
• Will be submitted to ISO/TC 211 later at appropriate time
• OGC 17-087 Features and Geometries Part 1 Feature Models will be circulated The scope of the replacement seems to be much wider than that of ISO 19125-1, which is to "standardize names and geometric definitions for Types for Geometry" - concentrating on 'simple geometry'.

For example, the new OGC "Features and their Geometries: Part 1" would require a change to the General Feature Model (19109) to allow feature instances to 'belong to' more than one feature type. The current "Simple Features" model ignores several issues that have been, not surprisingly, ignored in implementations.

The first issue looks at the use of UML as a language for the feature data model, while the application in the field is likely to be based on dynamic object models which are ignored the current "feature model". This generally ignored "big data" applications which date after the previous models. The extensions use taxonomies and ontologies (which include UML models). This sort of needed extension is to be expected, considering that the primary draft of Simple Features dates to the late 1990s. Things change over time.

The second volume looks at the non-Euclidean nature of surfaces used in geodesy (ellipsoids - intrinsic geodesy). Currently, only the GIS applications suppliers take care of the issue for non-Euclidean surfaces. Most of the "open source" applications use Pythagorean metrics or remain quiet on the topic.

The new series of documents replacing Simple Features simply dropped the word "simple" and dealt with the same issues and with those that Simple Features ignored. The Geodesy literature refers to it as intrinsic geometry, which in mathematics refers to the use of Differential Geometry. Hence, the new Part highlights the Riemannian metrics which are derived from surfaces embedded in 3D Cartesian space (called the First Fundamental Form). The alternate mechanism is to do the calculation in 3D Euclidean projected down to the ellipsoid, which has issued more complicated than Riemannian integrals. In fact, the integrals still need to be used but transformed into XYZ coordinates. Keeping the curves in XYZ to stay on the surface of an ellipsoid is difficult but necessary to match the correct values for distance (along a curve feature), or area (of a surface feature). Transforming to phi, lambda is significantly more efficient.

The integrals of the Riemannian metric can easily be approximated by iterative summations of a numeric integral approximation.

Simple Features (ISO 19125-1:2004) was carefully written to only include only topological query operations which do not require the heavy mathematics need for the metric operations on the ellipsoid.

The current plan will replace 19125, but it will also create a consistent basis for valid metrics (either 3D Euclidean or Riemannian ellipsoidal).

If users wish to use Euclidean metrics, then they will need to stick to large scale maps covering a smaller area where an "Engineering-drawing" model works well enough.

Part 3, yet to be started, will add the geometries in 19125-1:2004 As discussed in Maribor, OGC Features & Geometries - Part 1 is a really useful piece of work, introducing non-relational databases & the 'open world' model.

This could have far ranging consequences across a range of TC211 standards, if we choose to handle it that way. Currently there is little 'standardisation' about feature instances. We could amend/extend the General Feature Model (19109), or introduce something 'above that', of which the GFM is the 'relational profile'.

We probably need a new conceptual schema language (19103) - if such a thing is actually relevant in this world. Here, I'm not sure that 'best practice' has emerged?

There are knock on effects ranging from Data Quality (how to check conceptual consistency, when the 'real world' is the arbiter?) to encoding standards.

Personally, I would see 19125 as 'effectively untouched', the simple relational profile of all this new stuff! An "effectively untouched" Simple Features has issues mostly based on geometry.

ISO 19125 is mute on geographic issued in geometry, in which case SF pretty much sticks to Engineering Coordinate Systems or "small area" maps where the area is sufficiently small that errors introduced by the Earth's curvature, probably something like a city map (probably not greater Los Angeles, 109 mi. North to South). A local spherical approximation can be used but eventually, the full intrinsic geometry of the ellipsoid (maybe later, the geoid) needs to be addressed.

Simple Features were and still are limited in what it can do. Spatial operators (based on Engenhofer) are all strictly topological. Looking from ground level 5 feet 7 inches (2.9 meters) can see about 2.9 miles (4.7 km). At a 100 ft (30 m) lookout tower, can see 12.2 miles or 19.6 km. See "https://www.quora.com/When-I-look-out-into-the-ocean-how-far-away-is-the-horizon-How-much-of-the-ocean-can-I-actually-see".

Which indicates that a Simple Feature data set would probably be okay under the 12-mile (US Territorial) limit; where the curvature of the earth has its reasonable visual limits for a line of sight.

[gbuehler]

To be discussed at OGC meeting in Wellington (Dec 2017). Replacements to be consistent with SQL-MM. New name and number to be decided. From the Wellington JAG report (N4706):

• To be developed in a multi part series of standards
• New approach based on features and their description
• Will be submitted to ISO/TC 211 later at appropriate time
• OGC 17-087 Features and Geometries Part 1 Feature Models will be circulated The scope of the replacement seems to be much wider than that of ISO 19125-1, which is to "standardize names and geometric definitions for Types for Geometry" - concentrating on 'simple geometry'.

For example, the new OGC "Features and their Geometries: Part 1" would require a change to the General Feature Model (19109) to allow feature instances to 'belong to' more than one feature type. The current "Simple Features" model ignores several issues that have been, not surprisingly, ignored in implementations.

The first issue looks at the use of UML as a language for the feature data model, while the application in the field is likely to be based on dynamic object models which are ignored the current "feature model". This generally ignored "big data" applications which date after the previous models. The extensions use taxonomies and ontologies (which include UML models). This sort of needed extension is to be expected, considering that the primary draft of Simple Features dates to the late 1990s. Things change over time.

The second volume looks at the non-Euclidean nature of surfaces used in geodesy (ellipsoids - intrinsic geodesy). Currently, only the GIS applications suppliers take care of the issue for non-Euclidean surfaces. Most of the "open source" applications use Pythagorean metrics or remain quiet on the topic.

The new series of documents replacing Simple Features simply dropped the word "simple" and dealt with the same issues and with those that Simple Features ignored. The Geodesy literature refers to it as intrinsic geometry, which in mathematics refers to the use of Differential Geometry. Hence, the new Part highlights the Riemannian metrics which are derived from surfaces embedded in 3D Cartesian space (called the First Fundamental Form). The alternate mechanism is to do the calculation in 3D Euclidean projected down to the ellipsoid, which has issued more complicated than Riemannian integrals. In fact, the integrals still need to be used but transformed into XYZ coordinates. Keeping the curves in XYZ to stay on the surface of an ellipsoid is difficult but necessary to match the correct values for distance (along a curve feature), or area (of a surface feature). Transforming to phi, lambda is significantly more efficient.

The integrals of the Riemannian metric can easily be approximated by iterative summations of a numeric integral approximation.

Simple Features (ISO 19125-1:2004) was carefully written to only include only topological query operations which do not require the heavy mathematics need for the metric operations on the ellipsoid.

The current plan will replace 19125, but it will also create a consistent basis for valid metrics (either 3D Euclidean or Riemannian ellipsoidal).

If users wish to use Euclidean metrics, then they will need to stick to large scale maps covering a smaller area where an "Engineering-drawing" model works well enough.

Part 3, yet to be started, will add the geometries in 19125-1:2004 As discussed in Maribor, OGC Features & Geometries - Part 1 is a really useful piece of work, introducing non-relational databases & the 'open world' model.

This could have far ranging consequences across a range of TC211 standards, if we choose to handle it that way. Currently there is little 'standardisation' about feature instances. We could amend/extend the General Feature Model (19109), or introduce something 'above that', of which the GFM is the 'relational profile'.

We probably need a new conceptual schema language (19103) - if such a thing is actually relevant in this world. Here, I'm not sure that 'best practice' has emerged?

There are knock on effects ranging from Data Quality (how to check conceptual consistency, when the 'real world' is the arbiter?) to encoding standards.

Personally, I would see 19125 as 'effectively untouched', the simple relational profile of all this new stuff! An "effectively untouched" Simple Features has issues mostly based on geometry.

ISO 19125 is mute on geographic issued in geometry, in which case SF pretty much sticks to Engineering Coordinate Systems or "small area" maps where the area is sufficiently small that errors introduced by the Earth's curvature, probably something like a city map (probably not greater Los Angeles, 109 mi. North to South). A local spherical approximation can be used but eventually, the full intrinsic geometry of the ellipsoid (maybe later, the geoid) needs to be addressed.

Simple Features were and still are limited in what it can do. Spatial operators (based on Engenhofer) are all strictly topological. Looking from ground level 5 feet 7 inches (2.9 meters) can see about 2.9 miles (4.7 km). At a 100 ft (30 m) lookout tower, can see 12.2 miles or 19.6 km. See "https://www.quora.com/When-I-look-out-into-the-ocean-how-far-away-is-the-horizon-How-much-of-the-ocean-can-I-actually-see".

Which indicates that a Simple Feature data set would probably be okay under the 12-mile (US Territorial) limit; where the curvature of the earth has its reasonable visual limits for a line of sight.