Proposal for feature model review

[Martin Desruisseaux <de...@apache.org>]

Hello all

The org.opengis.feature package is currently in the "pending" part of
GeoAPI [1]. I would like to bring it to the normative part and submit it
to an OGC meeting, however I feel that the package needs some review
(some of those issue are reported in JIRA [2]):

  * An ISO 19109 (Rules for application schema) draft specification help
    to understand better the ISO model.
  * The distinction between XXXDescriptor and XXXType does not exist in
    ISO model, and its usefulness in practice is questionable. We could
    let such split to implementators who want it.
  * The distinction between SimpleFeature and ComplexFeature does not
    exist in ISO model, and is not the kind of specialization given by
    ISO examples. While a few convenience methods are useful, the need
    for a specialized SimpleFeature interface is questionable. We could
    leave this choice to implementors.
  * Defining Feature as a subtype of Attribute is a departure from ISO
    model and an unusual approach (would be similar to defining Class as
    a specialization of Field).
  * Defining IllegalAttributeException as a subclass of
    IllegalArgumentException is inconsistent with the way it is actually
    used (IllegalAttributeException is thrown by the validate() method
    for reporting an illegal /state/).
  * Design of current org.opengis.feature package seems tied to the GML
    representation (e.g. ComplexType.isInline()), while a current trend
    is the raise in popularity of alternative formats (e.g. JSON). I
    would like to avoid XML-specific properties from the API, and leave
    them to implementors.


I would like to define a much simpler feature model, very close to ISO
19109, with very few convenience methods (I envision only 2) and let all
remaining extensions or conveniences to implementors. Would the
following plan be okay?

  * Rename current org.opengis.feature package as
    org.opengis.feature.legacy.
  * Create a new org.opengis.feature package in the "normative" part of
    GeoAPI very close to the UML in figure 5 of ISO 19109:2013.
  * Discuss this model in a next OGC meeting (in addition to this
    mailing list).
  * New feature interfaces would be part of GeoAPI 3.1 if approved by
    OGC vote process (date to be determined).


What do peoples think?

    Martin

[1]
http://www.geoapi.org/snapshot/pending/org/opengis/feature/package-summary.html
[2] http://jira.codehaus.org/browse/GEO-229