[Requests] Comments to 09-146r3, OGC® Coverage Implementation Schema 1.1

Dimitri Sarafinof Dimitri.Sarafinof at ign.fr
Fri Jan 15 03:08:11 EST 2016


PART A

1. Evaluator:
        Dimitri Sarafinof - Dimitri.Sarafinof at ign.fr, Emmanuel Devys - Emmanuel.Devys at ign.fr

2. Submission: [09-146r3, OGC® Coverage Implementation Schema 1.1]

Note: This is not supposed to be an extensive review, which appears very difficult within a reasonable time-resource and due to the lack of the "global view" which the base UML model could have provided.

PART B

1. Requirement: General / Relationship with 19123

2. Implementation Specification Section number: General

3. Criticality: Major

4. Comments/justifications for changes: This specification is adding terms / concepts / classes of interest such as discrete-pointcloud (based on MultiPointcoverage) or discrete-mesh (grouping MultiCurve/Surface/Solid) or coverage-partitioning. It is also redefining / restructuring the structure of Coverage model and typology of classes of 19123 as well as the one adjusted in GMLCOV, and introducing classes of interest (such as SensorModelcoverage).
As a result, this specification appears as a "2 in 1" specification, with significant adjustments of 19123 (AS Topic 6), and revised UML models. Annex B provides only a partial view on this CIS coverage model.

Consequently, this specification appears as uneasy to apprehend and its relationship with 19123 need to be clarified.
1- Either by clarifying which scope of 19123 is covered, what is an extension, what is modified, with a clear mapping for terms / concepts and base classes of the model.
2- Or by providing 2 specifications instead of this "2 in 1" :
- a revised AS Topic 6 (19123) Coverage model, with adjustment of terminology with corresponding UML model
- a simplified CIS Implementation schema focused on implementation and encoding.
... or by any appropriate method that the editors would prefer.

The second solution is highly recommended. The revision of AS Topic 6 (19123) on the basis of CIS1.1 Coverage Implementation schema, which becomes necessary due to the gaps between 19123 and CIS, by some "backward or reverse engineering" appears as weird and difficult, with the current lack of the associated CIS UML model. This is even more important in the context of a document that is to be submitted to TC211 as 19123-2.

In any case, the availability of the CIS UML model is necessary in order to be able to apprehend, endorse and validate this new Coverage model (as far as 19123 or GMLCOV are the current references) and its XML schema artefacts provided in the submitted package.

Subsequently, it is believed - for clarity of users / implementers - that this Coverage schema should be clarified as some new version 2, instead of a simple revision of 09-146r2 in 09-146r3 as CIS1.1. It seems that backward compatibility with GMLCOV is not achieved (see further comment), so why minimizing this significant evolution of Coverage schema, which by the way is providing promising new features.


PART B

1. Requirement: General

2. Implementation Specification Section number: 5- Terms and definitions

3. Criticality: Major

4. Comments/justifications for changes: consistency with AS Topic 6 / ISO 19123 + internal consistency
1- Issues:
- Terminology issues: introduction of news terms in a specification that is supposed to be an implementation schema:
regular / irregular / distorted grid, Transformation grid, with no reuse nor maping with 19123 terms.

It is also recalled that 19123 terms (for coverages) are AS TOPIC 6 and ISO 19123 terms, in TC211 terminology, and that the change of terms need to be agreed with TC211, and show the mapping / relationship with 19123 terminology.
This is even more required for RectifiedGridcoverage (widely implemented), which are now (if my understanding is correct) a GeneralGridCoverage with a Grid geometry with RegularAxis.

2- Additional terms used in the specification should be added: partition, Sensor model.
3- Relationship between regular / irregular / distorted and Transformation grid with artefacts in the model should be clarified (i.e regular/irregular/Distorted/Transformation Axis).

Note: It is understood that the proposal of CIS model terms appears to be an improvement over the AS Topic 6 / 19123 terms. However, if this specification is supposed to be an implementation schema, the terminology for the coverage model should be defined in the proper specification, which is AS or 19123 revision. If kept as is, this terminology gap will be a difficult issue to handle in the Coverage communities and Coverage-based standards.


PART B

1. Requirement: General

2. Implementation Specification Section number: General

3. Criticality: Major

4. Comments/justifications for changes: CIS 1.1 is presented as a backward compatible solution of GMLCOV 1.0, which is not considering following facts :
                a) change of name GMLCOV => CIS
                b) change of namespace
                c) change of schemas (CIS schemas do not rely on GMLCOV schemas)
Ensure real backward-compatibility or make it a CIS 2.0.

Note : a simple example of this issue is trying to make a GMLCOV example, a grid Coverage (http://schemas.opengis.net/gmlcov/1.0/examples/exampleRectifiedGridCoverage-1.xml) compliant to CIS 1.1.
Changing a gmlcov:GridCoverage to cis:GridCoverage does not work as the definition as changed (contradictory with chapter 1.2.3)


PART B

1. Requirement: General

2. Implementation Specification Section number: General

3. Criticality: Major

4. Comments/justifications for changes: The proposed standard needs to be reworked to ensure compatibility and use with existing OGC standards (or with minor alignments).
Compatibility with existing OGC standards using GMLCOV 1.0 is not ensured :
                a) use of WCS 2.0 with CIS 1.1 seems impossible as WCS 2.0 rely on GMLCOV 1.0 (see requirements and schemas). For example wcsDescribeCoverage.xsd uses gmlcov:rangeType and gmlcov:metadata; CIS redefines them both!
                b) use of GMLJP2 2.0 with CIS 1.1 seems also impossible. A GMLJP2CoverageCollection is a subtype of gmlcov:AbstractCoverage and GMLJP2 schema heavily rely on GMLCOV schemas

CIS (as presented now) would require to update both WCS 2.0 and GMLJP2 2.0 standards.


PART B

1. Requirement: General - consistency between UML models (informative) and XML schemas

2. Implementation Specification Section number: General

3. Criticality: Major

4. Comments/justifications for changes: models in annex B and schemas provided are to be checked for consistency. The discrepancies increase the difficulty to apprehend this specification. examples : CIS::RegularAxis model has an origin defined as a string (!?) in UML model and nothing corresponding in XSD.


PART B

1. Requirement: Consider the alignment of specification with model and schema for Discrete-PointCloud

2. Implementation Specification Section number: 11 - Discrete-PointCloud

3. Criticality: Minor

4. Comments/justifications for changes: This class seems (if this is correct) to be a renaming of the MultiPointCoverage. It does not appear in schema nor in UML model (in fact just as a potential package). What is the interest in this discrepancy between class name in this specification and lack of evidence in schema?


PART B

1. Requirement: Consider the alignment of description and UML model and with the definition provided in 5.6

2. Implementation Specification Section number: 12 - Discrete-mesh

3. Criticality: Major (though impact is limited, as mesh has no evidence of implementation in schema)

4. Comments/justifications for changes: 5.6 gives an acceptable definition for mesh (though the impact of "respectively" is not crystal clear. The descriptive text in 12.1 seems to contradict this (extending class discrete-pointcloud), which does not seem correct ; this is (unfortunately) illustrated in figure 12, which include MultiPointCoverage. This does not seem to conform to commonly-accepted scope of mesh, and is confusing. However, it appears that mesh has no evidence of implementation in schema - so this definition appears totally "abstract".


PART B

1. Requirement: Consider the interest of Coverage partition / partitioning concept / class + align specification / UML and XSD.

2. Implementation Specification Section number: 16 - Coverage-partitioning

3. Criticality: Major

4. Comments/justifications for changes:
- what is the difference with a Coverage collection ? Why define a new structure instead of using the existing "collection" mechanism ?
- class "coverage-partitioning" is used in the specification, but this class name does not appear in UML nor in any XSD element. Consider to harmonize.


PART B

1. Requirement: How to handle duplication of envelope in CIS::AbstractGridCoverage (according to UML models) and (deprecated) CIS::GridCoverage and CIS::RectifiedGridCoverage ?

2. Implementation Specification Section number: 8.3.3 RectifiedGridcoverage + Annex B + annexed schemas

3. Criticality: Major

4. Comments/justifications for changes: According to specification and UML models, CIS::AbstractGridCoverage has now an envelope attribute (CIS::EnvelopeByAxis); CIS::GridCoverage and CIS::RectifiedGridCoverage have aGML::Envelope. This (apparent) duplication is to be clarified with some explanation and guidance, as RectifiedGridCoverage is presumably the dominant existing type of coverage data produced nowadays.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengeospatial.org/pipermail/requests/attachments/20160115/a7623daf/attachment-0001.html>


More information about the Requests mailing list