W3C home > Mailing lists > Public > www-rdf-dspace@w3.org > June 2003

Re: Relevant Technologies Document: Application Profiles

From: MacKenzie Smith <kenzie@MIT.EDU>
Date: Fri, 27 Jun 2003 14:40:08 -0400
Message-Id: <5.0.2.1.2.20030627142537.037f9f48@hesiod>
To: www-rdf-dspace@w3.org

I want to second what Dan said about this -- I've worked on application
profiles for a number of schemas, including a new one for METS which
actually has it's own schema
(http://www.loc.gov/standards/mets/profile_docs/mets.profile.v1-0.xsd)
The content of these profiles is almost always for humans and is
application rules for the schema as it applies in a particular context.

Not to say that machine-interpretable application profiles aren't useful
for some purposes (e.g. tailoring a UI) but, as Dan points out, this can
often be handled by the schema language directly...

P.S. And speaking of Biomedical Image Archives, the one which spawned
our own SIMILE use case (2.4 Biomedical Images) is on display at
http://ted.hul.harvard.edu:8080/ted/deliver/home?_collection=bil
If you go to a full record display (e.g. browse the term "macrophages")
you can see the underlying descriptive metadata schema and how
different it is than the usual approach (starts with Project, includes
instrumentation groupings, etc.). That system was built on the
Tamino XML database.

MacKenzie

At 01:34 PM 6/23/2003 -0400, Dan Brickley wrote:
>Just to butt in.... :)
>
>One reason 'application profile' emerged as a phrase that Dublin Core folks
>were using was the perceived problem that schema languages, DTDs, ontology
>languages and suchlike didn't provide a formal way of capturing deployment
>practices. Particularly at the RDFish end of the spectrum, which defines
>terminology rather than file formats, we found we didn't have a way to say
>(in a machine format) things like:
>
>(fictional, but loosly based on what I remember from my brisbio work @ILRT)
>
>  "In ILRT's Biomedical Image Archive we use Dublin Core title/description
>properties. Each title should begin with a capital letter, and end with a
>full stop. Titles are mandatory in the biomed database. Descriptions are
>optional, but should not exceed 1000 characters. The values of dc:subject
>are drawn from the MeSH classification system. We write dc:date in ISO8601
>and use the date that the image was digitized."(etc etc).
>
>In other words, there are many things about my application, and its
>structuring and organisation of data, which are not captured by the general
>schema/vocabulary description for Dublin Core, MeSH, etc.
>
>Syntax-oriented techniques, DTDs, XML Schema etc can go part of the way there,
>since you can often write a syntactic schema which captures some
>broad-brush aspects of this. The XML Schematron work was imho the best in
>that vein, as it decouples data checking from vocabulary design and file
>format structure.
>
>Anyway, people basically wanted to generalise about their use of RDF/XML
>vocabularies, and in particular the ways in which they could combine these
>to meet some application need, and were noting that this didn't seem to be
>what RDFS, OWL etc were designed to do.
>
>I'm currently of the view that prose is under-rated here, and that a
>human-oriented narrative (perhaps + test cases, examples) is a better way to
>document such practices than seeking a machine-friendly 'application profile'
>language.
>
>imho etc.,
>
>Dan

MacKenzie Smith
Associate Director for Technology
MIT Libraries
Building 14S-208
77 Massachusetts Avenue
Cambridge, MA  02139
(617)253-8184
kenzie@mit.edu
Received on Friday, 27 June 2003 14:40:17 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 13:35:25 EDT