DCMI Architecture Forum - Report - Telecon 2010-08-19, Thursday

Agenda: http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1008&L=DC-ARCHITECTURE&P=3323

Expected: Tom Baker (chair), Pete Johnston, Mikael Nilsson, Dan Brickley, Jon Phipps

----------------------------------------------------------------------
RDFa 1.1 profile

   Pete: Term mappings are possible - "title" mapped to
   "dcterms:title", etc. Also prefix mappings: if you use
   CURIEs, you can put the prefix mappings into the document
   or put them into a profile and just point to the profile.
   People don't need to put everything into the instance
   documents.  The scope of an RDFa profile does not
   necessarily need to be same as scope of a namespace -
   one profile can cite prefixes and term mappings for
   different namespaces.  Can be used to bundle a set of
   commonly used namespaces.  I have no strong opinion
   on whether DCMI should do this.  Instead of doing in a
   vacuum, could think of concrete use case.  For example,
   when exposing DCMI Metadata Terms, we are experimenting
   with RDFa, using a concrete set of terms in combination.
   As part of that work, we could create a profile covering
   the set of terms used to describe DCMI Metadata Terms,
   including references to terms in other namespaces.  From a
   social perspective, that would be a good thing to do.

   Tom: Bookmarking as action for the medium term.  DCAM
   discussion currently a higher priority.

----------------------------------------------------------------------
DCMI feedback on HTML Microdata
   Proposed feedback to [log in to unmask]:
   https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1008&L=DC-USAGE&P=52

   Nods of approval from participants on the call.  No red
   flags. Tom will review and post.

----------------------------------------------------------------------
DCAM: Past, present, future 
   http://dublincore.org/architecturewiki/DcamInContext

   Dan: DCAM importance perhaps more sociological than
   technical.  Dates from an era in which RDF was not
   deployed and looked risky.  DCAM was a compromise: given
   a community model, "these are the things in RDF of value
   for us".  Nowadays, with SKOS and Linked Data, RDF is
   looking healthier.  We could take the position that it
   was a document "from its time".

   Mikael: What DCAM does that is important to DC community:
   it provides a set of typical ways of expressing simple
   statements.  RDF itself is very pure and abstract; it can
   be hard to see how to organize triples.  DCAM - different
   view on triples.  Designed for specific communities with
   heavy resources, such as libraries, with lots of properties
   describing resources.  (As opposed to other communities
   with lighter metadata.)  Turn to DCAM: "how do I encode
   my metadata"?  Even if they do not follow DCAM in detail,
   they may follow its patterns.  If DCAM were removed, would
   still be a need for 20-30 typical patterns.  These patterns
   could be linked to the syntaxes like RDFa and RDF/XML.

   Jon: I challenge the assumption that DCAM hasn't had takeup.
   It has in a social context, in framing the problem.
   And people use it training.  Agree it could be helpful
   to de-emphasize DCMI-specific language.  See it as a
   transitional object between the XML model of the world
   and RDF.  Problem with DCAM: there is not always a clean
   distinction btw syntax and semantics.

   Pete: Broadly agree with Dan.  It was created at
   a particular point in time to perform that social
   function. Agree with Jon, could in theory be seen as
   transitional thing btw classical Dublin Core world and
   RDF world.  Not clear in own mind whether this is the time
   to push the bridge into history and move to a pattern-based
   approach or whether to try to maintain it as bridging device
   (i.e., continue current course).  Both approaches possible.
   Maybe continue to maintain and couch as purely syntactic.
   Or deprecate and replace with series of patterns.  Want to
   make cleaner separation btw DCAM abstract syntax, on one
   hand, and DSP/Application Profiles.  In RDF context, could
   offer that sort of functionality using query patterns.
   Also attracted to Dan's proposition (let go and move into RDF
   world).

   Tom: We need to go into the Pittsburgh meeting with
   a set of possible outcomes, prioritized as to which we
   see as most desirable -- a straw-man recommendation to
   be supported or refuted.

   Pete: As a next step, need to get options onto the mailing
   list for discussion.

   Mikael: DCAM maybe not best as a normative model, which it
   is today, but an informative document -- for documenting
   patterns.  That ties in with Pittsburgh meeting: the
   important point is to find what problems people actually
   are having and need to solve regarding application profiles
   in the RDF, Linked Data, and DC worlds.  The very existence
   of DCAM provides a point of reference: a way to structure
   your metadata (with the additional part, DSP).

   Tom: When writing [1], didn't seem to make sense to discuss
   DCAM without discussing DSP which, in a sense, provides
   the rationale for having a DCAM.  Everyone please give my
   deliberately syntax-oriented summaries of DCAM and DSP a
   closer read to see if they are both accurate and clearly
   understandable.  Tried to characterize both without using
   jargon -- a fine line, because in so doing I didn't want to
   create yet more jargon.

----------------------------------------------------------------------
Joint meeting, DCMI Architecture Forum and W3C Library Linked Data Incubator Group
   http://www.asis.org/Conferences/DC2010/program-sessions.html#jointmeeting

   Title: Application Profiles for Linked Data: Models & Requirements (Parts 1 & 2)
   Convenors: Tom Baker, Emmanuelle Bermes, Antoine Isaac
   Sponsors: DCMI Architecture Forum and W3C Library Linked Data Incubator Group
   Date/Time: Friday 22 October, 2:00-5:30pm
   Abstract: Application Profiles for Linked Data: models and requirements"

   This two-part meeting will look at current approaches to
   application profiles -- methods for documenting the content
   of descriptive metadata to promote the design of compatible
   metadata applications and maximize the coherence of metadata
   in a Linked Data environment. Starting with a review of
   the approach based on the DCMI Abstract Model, including
   the Description Set Profile constraint language and
   Singapore Framework for Dublin Core Application Profiles,
   the meeting will also consider other emerging approaches
   to specifying and documenting metadata patterns for use
   in Linked Data. By taking a fresh look at requirements
   in a rapidly evolving environment, this meeting aims at
   identifying and prioritizing areas where future work may
   be needed.

----------------------------------------------------------------------
Next steps

-- Tom to post comments on HTML Microdata.

-- On dc-architecture:
   Review the wiki draft [1] for correctness and clarity.
   Define range of possible courses of action, moving forward.
   Encourage discussion of which course to follow.

-- On another telecon in second half of September:
   Prioritize scenarios, decide on course of action to recommend.
   Approve wiki draft for circulation, e.g., to LLD list.

[1] http://dublincore.org/architecturewiki/DcamInContext