[VM] VM Task Force progress report

Dear all,

In the VM telecon last week, I had an action to prepare
excerpts from previous telecons for inclusion in the new
Editor's Draft [1].

This has turned into a more complete progress report describing
how the focus of VM has evolved over the past few months and
what we still hope to achieve by the end of January.

My intent was to provide here a one-stop source of all the
current links and issues.

Tom

[1] http://www.w3.org/2005/11/15-vmtf-minutes.html#action06

-----

SWBPD Vocab Management Task Force - Progress report - 2005-11-22

1. VM activity thru mid-September 2005

In 2004 and 2005 the VM Task Force worked on a draft "Managing
a Vocabulary for the Semantic Web - Best Practice" [14].  In
May 2005, this draft was superseded by the more tightly-focused
"Basic Principles for Managing an RDF Vocabulary" [5].

As described below, both of these drafts have now been
superseded by the even-more-tightly-focused "Configuring
Apache HTTP Server for RDFS/OWL Ontologies Cookbook" [12] --
here known as the Apache Configuration Cookbook.

The older superseded draft [14] is still listed under "Other
Editor's Drafts on the BPD Web page and should probably now
be replaced with a link to [12].  The description of the
Task Force at [13] is also out of date and could perhaps be
replaced with a document based on this posting.


2. VM activity mid-September 2005 thru end-January 2006

In September, at an information face-to-face meeting at
DC-2005 in Madrid, some members of the TF agreed to focus
on a practical solution to the question: What does one get
when dereferencing the URI of a property?  

After considering alternative solutions such as embedded RDF
(XHMTL 2.0) and RDF/A syntax, the TF agreed [1] to focus on
content negotiation [4].  Alistair proposed the following
requirements [16]:

     1) If a person tries to dereference the URI of a class or
        property (i.e. via a Web browser), they end up at
        the relevant bit of human-readable documentation.

     2) If a machine tries to dereference the URI of a class
        or property, they end up with a serialisation of a set of
        RDF statements describing that class or property, with
        a provenance that allows differentiation of different
        'versions' of an RDF schema/ontology.

     3) The whole thing complies with TAG resolution on
        httpRange-14 [6].

Alistair wrote up a set of examples [11] which has since then
morphed into the Editor's Draft of the Apache Configuration
Cookbook [12].  The hope is that the solutions outlined therein
can be embraced by FOAF, SKOS, Dublin Core, and other RDF
vocabularies, but for now the goal is to determine whether
the Apache configuration "recipes" are technically sound and
conform to relevant Web specifications.  

The TF would like to see this draft reviewed over the next
two months and use it as the basis for continued work in the
next round of charters.

Telecons are scheduled for Tuesdays at 1400 UTC on 22 November,
6 December, and 20 December.


3. Apache Configuration Cookbook

The Apache Configuration Cookbook [12] considers a range of
variables for dereferencing practice:

-- Vocabularies which use "hash" URIs (such as SKOS) as
   opposed to those which use "slash" URIs (such as FOAF
   and Dublin Core).

-- Maintainers who wish to meet "minimal" requirements
   (providing an RDF schema) as opposed to those who wish to
   meet "extended" requirements (also providing documentation
   as HTML Web pages).

-- Maintainers who wish to provide one single Web page
   as documentation for an entire vocabulary as opposed to
   those who wish to provide separate pages for individual
   terms (an attractive option for large vocabularies).

-- Maintainers who use PURL.ORG for their term URIs (see 
   below).


4. Issues under discussion

-- URIs based on PURL.ORG. DC, RSS, and vocab.org 
   use URIs based on purl.org, which currently responds
   to GET requests with 302 (Temporarily Moved).  The TAG
   decision on httpRange-14 [6] requires 303 (Redirect).
   This was discussed in a thread starting at [20].  In July,
   Dan asked TAG whether a 302 response on a purl.org URI
   would be acceptable [21].  As of November, Alistair has
   included purl.org scenarios in the Apache Configuration
   Cookbook [23].

-- Awareness of multiple representations.  If content
   negotiation is handled in the background and alternative
   representations of a vocabulary in RDF or HTML are served up
   to a user seemingly automatically -- on the basis of Apache
   configurations -- how would an interested user know which
   other representations were available? [2]  Should there
   be some way for a user to learn about other representations
   (e.g., via cross-references or an overview page)?

-- "Definitive RDF description".  Peter Patel-Schneider
   has questioned the focus on one "definitive RDF description"
   for each RDFS/OWL class, property, or individual --
   as opposed to an RDFS description, an OWL description,
   or multiple "definitive" descriptions [9].

-- Versioning and change management.  A thread on versioning
   and change management was started in July [17] with a draft
   by Alistair about "configuration management" for RDFS/OWL
   ontologies [10].  There was discussion on whether this
   should more properly be called "change management" or even
   "version documentation".  DCMI practice is described in
   [25].  The emerging consensus is that it is best to version
   the description of a property (i.e., the RDF statements
   about it), not the property itself.  I.e. each 'version' is
   a named graph; provenance information can then be used to
   distinguish between different descriptions of a property.
   Alistair notes that the issues of versioning and change
   management are coming to the fore in the OWL community,
   as evident at the OWL Futures Workshop at ISWC in Galway
   (see also [18] and [19]).

-- Provenance and URIs.  Provenance is supported by using
   the final URI from the chain of redirects as the name of
   the graph; different URIs represent different versions of a
   vocabulary.  Tom has noted that, in practice, "date-stamped"
   URIs are often used and suggests we explicitly acknowledge
   both that URI strings are in theory opaque and unparsable
   and that there are de-facto social conventions for using
   date stamps or version numbers [15].

-- rdfs:isDefinedBy.  Clarifying the dereferencing options
   provides an opportunity to clarify good practice for
   the use of rdfs:isDefinedBy [26].  Instead of relying on
   URI string manipulation in an attempt to heuristically
   locate a namespace, namespaces should be declared with
   rdfs:isDefinedBy.


REFERENCES

 [1] http://www.w3.org/2005/09/27-vmtf-minutes.html
 [2] http://www.w3.org/2005/10/28-vmtf-minutes.html
 [3] http://www.w3.org/2005/11/15-vmtf-minutes.html
 [4] http://www.w3.org/TR/webarch/#def-coneg
 [5] http://www.w3.org/2001/sw/BestPractices/VM/principles/20050705
 [6] http://lists.w3.org/Archives/Public/www-tag/2005Jun/0039.html
 [7] http://isegserv.itd.rl.ac.uk/VM/http-examples/
 [8] http://danbri.org/words/2005/10/25/142
 [9] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0075.html
[10] http://esw.w3.org/topic/ConfigurationManagement
[11] http://isegserv.itd.rl.ac.uk/VM/http-examples/
[12] http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/
[13] http://www.w3.org/2001/sw/BestPractices/VM/
[14] http://esw.w3.org/topic/VocabManagementNote
[15] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0078.html
[16] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Oct/0024.html
[17] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Jul/0016.html
[18] http://isegserv.itd.rl.ac.uk/cerif/lite
[19] http://www.co-ode.org/downloads/owldoc/co-ode-index.php
[20] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Aug/0037.html
[21] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Jul/0056.html
[23] http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe6
[24] http://lists.w3.org/Archives/Public/public-swbp-wg/2005Aug/0038.html
[25] ftp://ftp.cenorm.be/public/ws-mmi-dc/mmidc148.pdf
[26] http://www.w3.org/2000/01/rdf-schema#isDefinedBy
[27] http://www.w3.org/TR/rdf-schema/#ch_isdefinedby


APPENDIX: How some vocabularies currently handle dereferencing

-- FOAF vocabulary.  The FOAF vocabulary is named by the URI
   http://xmlns.com/foaf/0.1/ and has information available
   at that URI in both machine-friendly and human-friendly
   form. When that URI is dereferenced by HTTP, the default
   representation is currently HTML-based but includes an
   in-line RDF/XML description of the vocabulary.  This is
   described in more detail at [8].

-- MARC Relator vocabulary.  Library of Congress identifies
   MARC Relator terms with hash URIs (e.g.,
   http://www.loc.gov/loc.terms/relators/ILL).  Dereferencing
   this URI returns an HTML page showing the label and
   definition of the property in question.  Currently,
   property URIs do not dereference to the RDF schema
   (http://www.loc.gov/loc.terms/relators/dc-relators.xml).

-- Dublin Core.  DCMI identifies its terms with a
   hash URI (e.g.: http://purl.org/dc/elements/1.1/title).
   Dereferencing this URI gets the response code 302
   (Temporarily Moved) and a redirect to an RDF schema, e.g.,
   http://dublincore.org/2003/03/24/dces.  For example, the
   Purl server running at http://purl.org translates a GET
   request for http://purl.org/dc/elements/1.1/title into a GET
   request for http://dublincore.org/2003/03/24/dces#title.
   The http://dublincore.org server serves up that page with
   a status code of 200.


-- 
Dr. Thomas Baker                      baker@sub.uni-goettingen.de
SUB - Goettingen State                            +49-551-39-3883
and University Library                           +49-30-8109-9027
Papendiek 14, 37073 Göttingen

Received on Tuesday, 22 November 2005 11:28:22 UTC