RE: Namespace evolution.

Dear Roland,

Let's say there are three levels of change for metadata elements:
-- changes so small they are merely "editorial" and can be handled in 
   "errata" notes for an existing version;
-- semantic changes (substantive changes in definitions) that are 
   significant enough to emphasize with a new token and version number
   (whether at the namespace level or the element level);
-- semantic changes so fundamental that they imply a new element
   altogether.

We suggest that semantic changes on the second level are (almost by
definition) slight enough for many applications to ignore.  However,
applications that really need to be precise about semantics might use
date stamps to do so.  

In the paper, we suggest creating time-stamped tokens for each new
version change and linking the latest one at any given time to a
version-independent token.  The idea is that most people would use the
version-independent token, which would be precise enough for many
applications and would avoid having specifically to map between, say,
http://purl.org/dc/elements/1.0/title and
http://purl.org/dc/elements/1.1/title.

Since this is DC-specific I have copied this to the dc-registry list
and suggest we follow up there if necessary...

Cheers,
Tom


On Mon, 16 Oct 2000, schwaenzl wrote:
> there seems a loop hole in your proposal: 
> 
> In case you allow DEFINITIONS - that is semantics - to be changed, how can you 
> understand a specific record using such a name space? 
> How can you know the semantically validating definition of a tag in a record - 
> 
> Cheers
> rs
> 
> > On Mon, 16 Oct 2000, Sigfrid Lundberg, Lub NetLab wrote:
> > > On Mon, 16 Oct 2000, Jason Diamond wrote:
> > > > I just found this document, "Using Dublin Core in XML" [1], also a Working
> > > > Draft but much more recent, that uses http://purl.org/dc/elements/1.1/ as
> > > > the namespace. Thanks in advance to everyone who was going to point that
> > > > out. I still don't like the fact that it's version dependant.
> > > 
> > > Neither do I. Again, if you want to influence the development of these
> > > documents, turn to the DCMI itself. 
> > 
> > Shigeo Sugimoto, his colleagues, and I just finished a short paper
> > discussing some versioning requirements and alternative ways to version
> > the Dublin Core. We also come down in favor of version-independent
> > namespace and element names.  At the same time, we suggest some
> > mechanisms for helping developers or archivists specify or reconstruct
> > the exact state of namespaces and definitions as of a particular date.
> > 
> > We "very soon now" wanted to turn this paper into a strawman for
> > discussion on the mailing list of the DCMI Registry Working Group
> > (http://www.mailbase.ac.uk/lists/dc-registry/join.html) and welcome
> > anyone from this list to join.  I would be happy to share this paper
> > with anyone in the meantime.

_______________________________________________________________________________
Dr. Thomas Baker                                            Thomas.Baker@gmd.de
GMD Library
Schloss Birlinghoven                                           +49-2241-14-2352
53754 Sankt Augustin, Germany                              fax +49-2241-14-2619

Received on Monday, 16 October 2000 08:49:29 UTC