Hello all
The OGC meeting is now over. There is some points that attracted my
attention:
Metadata
-----------------------------------------------------------
Latest version of ISO 19115-1 has been officially released last week
[1]. We will need to update the GeoAPI interfaces and Apache SIS
implementation accordingly. I will create JIRA tasks for this work. This
revision contains no new mandatory elements, but some new optional
elements and a refactoring of some existing elements.
Latest version of ISO 19115-3 (formally ISO 19139) is under review. From
an Apache SIS point of view, roughly speaking ISO 19115-1 defines the
API of org.apache.sis.metadata.iso packages, while ISO 19115-3 defines
that JAXB bindings for XML (un)marshalling. A set of XSLT for
transforming ISO 19139 to 19115-3 will be provided by ISO, but I do not
know yet if the license terms would be compatible with Apache 2 license.
Systematic review of ISO 19115-2 (metadata extension for gridded data)
may take 2 or 3 years. Apache SIS would probably keep the current
classes unchanged in the main time. They are mostly the classes in the
org.apache.sis.iso.acquisition package.
Metadata in NetCDF files
-----------------------------------------------------------
The current NetCDF "classical" format (a.k.a. NetCDF 3) can store some
metadata in global attributes (using the CF convention), but this
provides a flat view of metadata. For some communities this is
sufficient. However there is an increasing need to organize metadata in
a more hierarchical fashion, for example using groups. The fact that
some peoples use NetCDF variables for simulating groups is probably a
symptom of this need. Groups are supported only in NetCDF 4 or HDF 4 and 5.
There is a need to establish a mapping between NetCDF attributes and ISO
19115 metadata. The current approach is to consider NetCDF classical
format as legacy and look forward by designing a mapping that use
groups. Then later, one could define a flattening mechanism for
simulating groups in a flat format like NetCDF classic model. In my
understanding, this work happen in the ESIP community [2].
Coordinate Reference System (CRS)
-----------------------------------------------------------
We had a reminder that EPSG database contains (since 2 years) the CRS
domain of validity as polygons. What we had before was areas of validity
specified as geographic bounding boxes. The polygon types are LinearRing
or a collection of LinearRings, expressed in a common reference system:
WGS84. Their precision is sub-kilometres. If I understood right, those
geometries are provided as Shapefiles. For Apache SIS, this would
introduce a small optional dependency from sis-referencing to
sis-shapefile module.
We have been reminded that the EPSG registry web site [3] provides
catalog services based on the ebRIM 3.0 specification. Users can make
standard HTML POST/GET queries and the registry returns objects wrapped
in GML version 3.2.1. Example:
http://epsg-registry.org/export.htm?gml=urn:ogc:def:crs:EPSG::4326
This work also for the above-cited polygons, example:
http://epsg-registry.org/export.htm?gml=urn:ogc:def:extent-polygon:EPSG::1262
As a side discussion, we noted that other sites exist, in particular
http://epsg.io/ (not officially related to EPSG despite its name) and
http://spatialreference.org/. But the Coordinate Reference System (CRS)
definitions provided by those sites - at least the ones in Well Known
Text (WKT) forms - are not conformant to EPSG definitions, especially
regarding axis order. Peoples are advised to use
http://epsg-registry.org/ as the authoritative source.
Well Known Text (WKT)
-----------------------------------------------------------
Time in OGC specifications is often represented by strings in the ISO
8601 syntax. This syntax is suitable for instants in calendar + clock
systems. It can also represent an interval of repetition (what we may
call "multi-instants"), but can not represent a period (except in the
particular case where the period matches exactly a calendar element like
a month or a year).
There is a need to represent Web Map Service (WMS) layers for more
complex time. For example sparse / incomplete time series (i.e. with
gaps), or the annual rainfall statistics in the 1918-2012 period. Those
temporal objects are hard to represent with ISO 8601. One possible
approach is to handle geometry of time as concepts similar to spatial
geometries. An "instant" in time is similar to a "point" in space, a
"multi-instants" is similar to a "multi-points", a "period" similar to a
"line", etc. Pushing further the idea, it may possible to define a WKT
format for temporal objects in a way similar than what has been done for
spatial geometries. This would include the definition of WKT
representation of temporal coordinate systems.
The alternatives are under exploration at OGC.
Miscellaneous:
-----------------------------------------------------------
I have been told that the Java virtual machine on Android is slow for
numerical intensive tasks. Some GIS products on Android perform their
map projection with C/C++ libraries for that reason. For Apache SIS, we
should have the flexibility to use either the native Java implementation
provided by SIS, or the Proj.4 wrappers [4], at user's choice.
There will be a join OGC / W3C meeting in London 5-6th March on linked
data (semantic web). The overall idea is that current world is
data-centric, and the goal is to move to a knowledge-centric world. By
adding formal semantics and context to linked data, for example using
the Resource Description Framework (RDF) model, it becomes
understandable by software.
A proposed changes in OGC standards policy is under discussion: if a
standard does not have implementation and no CITE test, then it would be
said "provisional".
Martin
[1]
http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=53798
[2] http://wiki.esipfed.org/
[3] http://epsg-registry.org/
[4] http://www.geoapi.org/geoapi-proj4/index.html
