[geotk] Report on OGC meeting

[Martin Desruisseaux]

Hello all

Last week was the 81th OGC meeting in Exeter (UK), and the 22th meeting 
I attended. I neglected to post short reports about the meetings in the 
past, and though I should start on this mailing list (focusing on aspect 
that are of particular interest to Geotk).

Metadata in NetCDF files
---------------------
We already know about a mapping between NetCDF attributes and ISO 19115 
metadata. This mapping is implemented in the 
org.geotoolkit.metadata.netcdf package. It was a somewhat informal 
mapping up to date, but if I'm understanding right it should be on the 
track for an OGC standard this year.

The above mapping can only be incomplete, because NetCDF and ISO 
metadata are quite different. I have learn at this OGC meeting that some 
peoples are working on a way to insert ISO 19115 metadata directly in 
NetCDF file, preserving the ISO terminology. They were some discussion 
about whatever such kind of work should be based on NetCDF 3 or 4 
formats. It seems that hierarchical metadata would fit more naturally in 
NetCDF 4 files because of the new capabilities offered by this format, 
but they were some objections in the room because of the widespread of 
NetCDF 3 formats and software working only with that format. NetCDF 4 is 
about one year old.

They were also a talk about a proposal for uncertainties, i.e. some 
statistical information like standard deviation associated to each 
sample values.


Referencing
----------
Some keys authors of ISO 19111 were back at this meeting. There is a 
proposal for defining a new WKT format for referencing, to be called WKT 
2. The reason is that current WKT format is inconsistent with ISO 19111 
and interpreted in different (usually incompatible) way by different 
implementors (see 
http://www.geotoolkit.org/apidocs/org/geotoolkit/io/wkt/Convention.html). The 
new WKT 2.0 format would be incompatible with WKT 1 (it is not possible 
to create a format which is both compatible with WKT 1 and consistent 
with ISO 19111), but parsers should be able to recognize automatically 
if they are parsing a version 1 or version 2 string.

I talked a bit about the GIGS tests 
(http://www.geoapi.org/geoapi-conformance/apidocs/org/opengis/test/referencing/gigs/package-summary.html), 
which raised some interest. One open question however is how to 
integrate with the CITE tests. I'm investigating this avenue.

A representative of the Web Coverage Service working group did a talk 
about the CRS issues they had. However some of their issues were not ISO 
19111 or EPSG database issues, but rather artefacts of external projects 
that they looked at and though that they were representative of CRS 
standards. For example they though that they were a vast amount of names 
for the same axis ("x", "long", "longitude", etc.). This is not true: 
the EPSG database has only one name for the above axis: "Geodetic 
longitude". All the observed variety are consequence of approximative 
conversions from EPSG database to WKT or Proj.4 definition files, etc. 
I'm amazed to see how widespread is the belief that some widely-used 
products are automatically good representative of standards. Sometime 
they are, but not always and we still have to check at the source.


Data preservation
---------------
An interesting talk about what is needed for improving the chances that 
data we are creating today will still be readable in 50 years. They made 
many recommendations. One of them was to avoid opaque storage systems 
like database, use well-known file formats instead. Database are good 
for production use, but not necessarily for long-term storage. This is a 
slight advantage (admittedly minor since we still use a database) for 
avoiding opaque binary objects in a database schema like the one 
provided in the Geotk "coverage-sql" module, or using only binary 
objects that we can recompute from information stored as primitive 
database types (this is our current approach).

     Martin