FW: SIMILE call re: DSpace history system, 25-Apr-03

-----Original Message-----
From: Tansley, Robert 
Sent: Friday, April 25, 2003 2:50 PM
To: Jason Kinner (jason_kinner@dynamicdigitalmedia.com); Eric Miller (em@w3.org)
Cc: Smathers, Kevin; Bass, Mick
Subject: SIMILE call re: DSpace history system, 25-Apr-03


Please let me know if I've missed or misrepresented anything.

SIMILE call re: DSpace history system, 25-Apr-03
Present: Eric, Jason, Kevin, Rob

Jason: Looked at DSpace code and has history system output Primarily looking at the data first, not the code.  Prepared to talk about existing data, and Harmony model.

Eric: Has an existing record. Has been transformed a little

Naming objects: current naming policy=www.dspace.org/(type of object)/numeric subdivisions--needs documenting

HTML-rendered fields? use 'parsetype=literal'

namespaces? 
copyright text with nothing in it? not declare properties with no values hasPart= should have literal, not a URI.

variety of namespaces--dspace.org.  Use namespaces for things that resolve to more information.

There is a Harmony namespace, which resolves to the schema.  That's good practice. Web is underspecified in this regard.

relative URIs harmful, use absolute URIs

Questions: stability of Harmony work--use as-is or re-articulate? used small set.  Leaning towards re-articulation.

Jason: Can combine Harmony and DC to create enhanced schema, e.g. to annotate values with the format, e.g. date.

Chartered to create schema based on Harmony, and any other schemas out there

Modernise to latest version of Harmony
What is http://www.dspace.org/harmony/action/de06fa83fbfcdceba558655cdd4fb747?
DSpace actions might be better represented as classes in the schema (as opposed to text) class or instance of class?  Should be explicit in the schema.

Eric: Everything is untyped in history.  Don't know what's captured and what isn't

HTML encoding: XML bad at dealing with binary

Rob: It's about how you serialise the objects in DSpace.  Items have Dublin Core, everything else just has database columns

Eric; It's about how much of the processing must be done in the application. Escape the literals if you have unvalidated, from database, best to encode, if you know it's well-formed, use parsetype=literal

Rob: Most data is just VARCHAR text in the DB, entered by end users

Jason: DSpace about different formats? Mistake to tie down to HTML?

Rob: Question is, how granular/specific can we get?  Most data in the system is just plain text

Eric: RDF core working group integrating XML schema datatyping work into RDF.  May be able to cast data in the database into those types

Jason: Minimally, data has to be annotated to be usable by diff systems.  Was thinking of using DC for this.

Discussion on various methods of representing data in RDF, fine-grained and validated vs. opaque text; former is good if you can achieve that, but with current DSpace UI, might need to resort to the latter which pushes some effort onto the applications using the data

Eric: I am often on IRC channel: freenode.net #rdfig


ACTIONS:

Rob: propose and set up SIMILE IRC channel

Rob:  Look into privacy concerns to do with making history data a useful dataset for the RDF community

DSpace URI naming conventions--talk to Mick?

Jason to propose better scheme for URIs


 Robert Tansley / Hewlett-Packard Laboratories / (+1) 617 551 7624

Received on Friday, 2 May 2003 14:07:10 UTC