Hello all
The OGC meeting was hold in Geneva form June 10 to 13. We lost the last
day because of French train strike, so this email is only about June 10
to 12. In this email, I try to explain better why an item is relevant to
Apache SIS with separated "Relevance to SIS" paragraphs.
Standardization process
------------------------------------------------------------
OGC is considering to classify their standards in 3 groups:
* Community specification, developed outside OGC but so widely used
that they are worth to be adopted as an official OGC position.
* Provisional OGC standard, which follow OGC policy (modularization,
etc.) but have no evidence of implementation and no CITE tests.
* Full OGC standards: same as provisional except that there is strong
evidence of implementation and maturity, and CITE test exists.
GeoAPI is concerned by the "Provisional" versus "Full" OGC standard. For
example GeoAPI provides its own test suite (the "geoapi-conformance"
module [1]) instead than the CITE tests. This situation exists because
of the particular GeoAPI nature compared to other OGC standard, since
GeoAPI is neither an encoding standard nor a web service standard.
Closer ties between CITE tests and "geoapi-conformance" is desirable,
but we would need volunteer help for that.
GeoAPI
------------------------------------------------------------
I posted a subset of GeoAPI slides below. The most important thing is
the Feature model proposal:
http://people.apache.org/~desruisseaux/GeoAPI/2014-06.pdf
See also GeoAPI snapshot javadoc [3]
We had about 7 peoples in the room. We got no comment yet on the feature
model. We got a comment that explicit support for circles is needed for
user location with a mobile device in GeoXACML 3.0. GML already has
Circle and CircleByCenterPoint, but the geometry interfaces do not yet
have a Circle interface. The next ISO 19107 revision may include
circles, in which case the GeoAPI interfaces will be updated accordingly.
*Relevance to Apache SIS:* GeoAPI interfaces are implemented directly by
Apache SIS. If there is any comments on the above Feature model, please
let us know!
Moving features
------------------------------------------------------------
The current scope of moving features is massive data exchange. Scope of
a future revision may be web service interface. The current scope has
two parts: Part I defines the XML core encoding while part II defines a
simple CSV format. The later is expected to be used most often. Those
formats can be seen as an implementation of "ISO 19141:2008 - Schema for
moving features" [2] model. The OGC Moving Features specification is
targeted for completion on December this year.
*Relevance to Apache SIS:* we have not yet checked if the current
Feature API proposal [3] is consistent with Moving Features. This is
another area where volunteer help would be welcome.
Data quality and user feedbacks
------------------------------------------------------------
Apache SIS provides an "org.apache.sis.metadata.iso.quality" package [4]
based on standard published in 2003. The new standard revision, ISO
19157, combines 3 older standards in a unified conceptual model. When
Apache SIS will be updated for that new standard, this is expected to
double the number of classes in the
"org.apache.sis.metadata.iso.quality" package.
There is an other standard under development, ISO 8000, which has a yet
wider scope. ISO 8000 seems to have about 150 parts (if I understood
right), which I think would be beyond the Apache SIS scope. If I
understand right, ISO 19157 (which Apache SIS would support) could be
seen as a subset of ISO 8000. However if anyone wanted full ISO 8000
support, I suspect that it could be the scope of a different Apache project.
An other data quality model is UncertML. UncertML provides some
probability distributions that can be used to express uncertainties in
positions, continue attributes, etc., but does not try to address (in my
understanding) other kind of values like confusion matrix, counting
number of conflicts, etc. They were some discussion about how to combine
UncertML with ISO 19157.
*Relevance to Apache SIS:* we will need to update the
"org.apache.sis.metadata.iso.quality" package, but make clear that the
scope is not full data quality in ISO 8000 sense.
User feedbacks
------------------------------------------------------------
"Data quality" (ISO 19157) can be seen as a kind of metadata (ISO
19115), and "user feedbacks" can be seen as a kind of "data quality".
The GeoViQua and CHARMe projects are experiments for complementing the
existing ISO 19157 elements with accumulated reports from data users.
The idea is to provide rankings like on the Amazon web site. This is a
different model than the current one:
* ISO 19157 is a "producer quality model" (quality recorded by the
producer).
* GeoViQua and CHARMeare are "consumer quality model" (attempts to
post-qualify the dataset).
* An other model mentioned but not discussed is "stakeholder quality
model".
However the "consumer quality model" raises the issue of users as a
source of error. This issue may be reported by the ISO 19157
"metaquality" (quality of the quality) element. One way to reduce the
problem is to iterate with the users. For example sending a request for
more information, asking same question in different ways, suggesting
that the user move to a different location, asking other users in the
area to confirm the observation, re-evaluating the quality of previous
observations as new information is made available.
The above-cited GeoViQua project proposes QualityML as an extension of
UncertainML. QualityML is a dictionary for quality metadata encoding.
This is not yet a standard, but an experiment.
*Relevance to Apache SIS:* we may need volunteer work for checking if
user feedbacks should be supported as an
"org.apache.sis.metadata.iso.quality" extension, maybe in an
experimental module.
GeoTIFF and NetCDF metadata
------------------------------------------------------------
A mapping from GeoTIFF tags to the "Coverage" model (a generalization of
rasters model) has been published [5]. A similar mapping from NetCDF to
"Coverage" model is under way. A draft exists, but there is still open
issues with OPeNDAP, especially their use of index for data access.
After the "NetCDF to coverage" binding, a next goal may be "ncML to GML"
(ncML is a XML encoding of NetCDF metadata.
*Relevance to Apache SIS:* once the "NetCDF to coverage" standard is
published, Apache SIS may need to revisit its NetCDF mapping
implementation [6].
The European Space Agency (ESA) is working on an extension of NetCDF
standard attributes for Earth Observation products [7]. This proposal is
compliant with CF-NetCDF and NetCDF-U (uncertainty convention). Final
version is planned for July, and standardization through OGC may come later.
*Relevance to Apache SIS:* the mapping of Earth Observation attributes
to ISO 19115 metadata is not yet clear to me. But if a volunteer is
willing to investigate, we should augment our NetCDF mapping
implementation [6] with those Earth Observation attributes.
GML in JPEG2000
------------------------------------------------------------
GMLJP2 version 2.0 is under work. The new specification will clarify
some GMLJP2 1.0 ambiguities (e.g. axis order), and may add support for
"georeferenceable" [8] images (not to be confused with "georeferenced"
images, which are already supported). They were also a request for
motion image annotations.
(Transactional) Web Coverage Service
------------------------------------------------------------
WCS-t contains 3 operations: insert, delete and update. The "update"
operation is the hard one to define. Work is still going on the later.
As a side-note, some peoples expressed the need to be able to determine
if a bounding box contains any non-null value before to request a
coverage (raster) in that area. The goal is to reduce the transfer size.
Some peoples are investigating JPEG 2000 Interactive Protocol (JPIP) for
streaming of coverages with reduced band width usage. An other aspect
under study is asynchronous access to WCS coverages via FTP or on DVD
(user noticed when ready).
Linked data
------------------------------------------------------------
While I'm not aware of an OGC standard specifically related to linked
data, OGC is thinking about this long-term tendency. The current World
Wide Web hyperlinks focus on "document about things". The Linked Data
approach rather focuses about thing themselves. This will require things
to have a unique and persistent identifiers.
*Relevance to Apache SIS:* SIS already tries to prepare itself with the
IdentifiedObject interface [9]. We do not know yet if the SIS approach
will work well, since I'm not aware of feedback yet.
Mobile location services (MLS), a.k.a. "Mass market"
------------------------------------------------------------
The "mass market" working group has been renamed Mobile Location
Services (MLS) domain working group.
*Relevance to Apache SIS:* we have an orphan "SIS for Android" branch
[10]. If we get a volunteer for maintaining that branch, we may need to
follow Mobile Location Services activity in order to apply relevant
recommendations to our branch.
Martin
[1] http://www.geoapi.org/geoapi-conformance/index.html
[2]
http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=41445
[3]
http://www.geoapi.org/snapshot/javadoc/org/opengis/feature/package-summary.html
[4]
http://sis.apache.org/apidocs/org/apache/sis/metadata/iso/quality/package-summary.html
[5] https://portal.opengeospatial.org/files/?artifact_id=54813
[6]
http://sis.apache.org/apidocs/org/apache/sis/storage/netcdf/AttributeNames.html
[7]
http://wiki.services.eoportal.org/tiki-index.php?page=Prod-Trees+Resources
[8]
http://www.geoapi.org/snapshot/javadoc/org/opengis/metadata/spatial/Georeferenceable.html
[9] http://sis.apache.org/apidocs/org/apache/sis/xml/IdentifiedObject.html
[10] https://svn.apache.org/repos/asf/sis/branches/Android/
