Enrico,
Thanks for sharing with us the document which I think shows good ideas of how to provide
"standard" the information, in terms both content and syntactic, of the
CI_OnlineResouce element.
Currently, our CSW server provides the same information as your client
supports, except that the linkage is a url to a GetCoverage, not
GetCapabilities, request. I think that a more specific protocol based on
OGC07-107r3, as suggested in your document, is better and will more clearly
tell the nature of the request URL. We'll revise our server to adopt your
suggestion and will let you know once it is done.
Our purpose of providing GetGoverage, rather than GetCapabilities, url is to allow
non-WCS compliant application software, such as GRADS, to directly download a specific
variable, because a netCDF files may include several variables. We noted that there are
possibly many, or theoretically infinite, GetCoverage requests for an offered coverage.
The URL provided in our server is a "nominal" coverage which includes default
bbox, resolution, and format.
I think that it might be better for a CSW server to provide several URLs, including a GetCapabilities url and several getCoverage
URLs. But the maximum occurrences of "linkage" and "protocol" are both 1 and therefore it seems not possible
to provide multiple URLs for a specific netCDF data set. In our server, we actually are not dealing with a netCDF data set but a
"variable". Therefore, if a user searches for a "precipitation" data set, our server finds a netCDF file
containing a "precipitation" variable and returns, in linkage, a getCoverage url for the precipitation variable,
without mention any other variables in the same netCDF file.
Finally, my discussion on this file vs variable is specific to the netCDF data
sets shown in Unidata's THREDDS catalog, with which our CSW is interacting.
Also we are considering our potential users who uses non-WCS software. For OGC
community, I think that a getCapabilities URL might more appropriate in terms
of providing more complete information on an offered coverage (data set).
Regards,
Wenli
PS: Chengfang is taking two-week off to finishing her dissertation. She may not actively participate discussion and working on the the implementation although she'll read her email.
----- Original Message -----
From: Enrico Boldrini <boldrini@xxxxxxxxxxx>
Date: Monday, September 15, 2008 5:35 am
Subject: [galeon] CI_OnlineResource conventions for WCS data access
Dear Chengfang,
at the moment our client is able to access data from the WCS
server
given the following template information:
<onLine>
<CI_OnlineResource>
<linkage>
<URL>http://motherlode.ucar.edu:8080/thredds/wcs/fmrc/NCEP/NAM/Alaska_11km/NCEP-NAM-Alaska_11km_best.ncd?
</URL>
</linkage>
<protocol>
<gco:CharacterString>OGC:WCS-1.0.0-get-coverage</gco:CharacterString>
</protocol>
<name>
<gco:CharacterString>Total_precipitation</gco:CharacterString>
</name>
</CI_OnlineResource>
</onLine>
Anyway we can modify our client to be able to support the format
that
your server provides.
In general we think it would be interesting to start a discussion
on the
access metadata topic, to be able to cope
at best with this part of ISO Metadata specification.
Therefore we have started writing a small document with a few
considerations and proposal.
We are posting it also to the Galeon list, thinking that it will
be of
interest.
We wait for feedback and contributions..!
Enrico