[ESP] Dublin Core question

[Ethan Davis]

Hi Eric, Luca,

One problem we've run into with DCMI-Period and most other time encoding schemes
is the lack of relative times. We deal with a lot of real-time data and have
archives that cover a certain number of days back from present. This kind of
thing is hard to represent in most time encoding schemes. The only metadata
encoding I'm aware of that deals with relative time is ADN (see below) which has
a timePast element.

Just a heads-up for anyone dealing with real-time data.

Ethan

(+) ADN is a metadata scheme developed by people at the Alexandria Digital
Library, DLESE (Digital Library for Earth Science Education, and NASA.
http://www.dlese.org/Metadata/adn-collaboration/

Luca Cinquini wrote:
> 
> Hi Eric,
>     here at NCAR we are actually experimenting with describing our high
> level search and discovery metadata in Dublin Core, since it was the
> only schema people could agree on - at least for the time being. It
> looks like the DC people finally agreed on an XML schema for encoding
> both unqualified and qualified Dublin Core elements, but as far as I
> understand there is yet no official recomendation for encoding the DCMI-Box
> and DCMI-Period information (i.e. spatial coverage and temporal coverage
> respectively). Hopefully that will come soon.
> 
> In relation to your questions:
> 
> o for DCMI Type, we use the value "Collection" when writing metadata
> that describes a whole set of files
>     (see definition at
> http://dublincore.org/documents/2003/03/04/dcmi-terms/)
> 
> o for identifier, we use a URI system based on a hierarchical relation
> between datasets: for example, "ucar.cgd.pcm.XX" is the root identifier for
> all files produced by run XX of the model PCM, run by the CGD people at
> UCAR. A cross-institutional URI system could start with the name of the
> authority assigning the identifier, for example "esp:ucar.cgd.pcm.XX"
> 
> o Coverage we encode with the DCMI-Box and DCMI-Period, for example:
> <dc:coverage>
> <dcterms:spatial>
> <box xmlns="http://dublincore.org/documents/2000/07/28/dcmi-box/"
> name="the whole earth">
> <northlimit>90</northlimit>
> <eastlimit>0</eastlimit>
> <southlimit>-90</southlimit>
> <westlimit>360</westlimit>
> </box>
> </dcterms:spatial>
> <dcterms:temporal>
> <period xmlns="http://dublincore.org/documents/2000/07/28/dcmi-period/">
> <start scheme="W3C-DTF">1983</start>
> <end scheme="W3C-DTF">2000</end>
> </period>
> </dcterms:temporal>
> </dc:coverage>
> (even if, as I said, this is non-offical XML....)
> 
> o Relation-isPartOf could be used to express containment relationship
> between datasets, even if we haven't done so as yet.
> 
> o Format we simply use a free-form value, since for example "netcdf" is
> not listed in the recomended list of MIME types....
> 
> Let me just conclude by saying that it would be great if this group
> would come up with a joint recomandation on how to use DC for high-level
> description of scientific data....
> 
> thanks, Luca
> 
> Eric Marshall wrote:
> 
> > What work has been done or discovered regarding how this community
> > will want to standardize how folks will describe datasets in Dublin
> > Core elements? I'm thinking of Type, 'a formal identification system'
> > for Identifier, Relation (and aggregation) and what would be useful
> > for Coverage and Format. Any thoughts?
> >
> >        Regards,
> >         Eric Marshall
> >
> > _______________________________________________
> > ESP mailing list
> > ESP@ucar.edu
> > http://mailman.ucar.edu/mailman/listinfo/esp
> 
> _______________________________________________
> ESP mailing list
> ESP@ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/esp

-- 
Ethan R. Davis                       Telephone: (303) 497-8155
Software Engineer                    Fax:       (303) 497-8690
UCAR Unidata Program Center          E-mail:    edavis@ucar.edu
P.O. Box 3000
Boulder, CO  80307-3000              http://www.unidata.ucar.edu/