- From: John S. Erickson <john.erickson@hp.com>
- Date: Wed, 7 May 2003 11:44:58 -0400
- To: "simile-w" <www-rdf-dspace@w3.org>
* Apologies to me HP mates for the dupe * Apologies for the TWiki markup John ---++++ Comments on Dspace History System: Descriptive Note (1 May 2003) _Revised 7 May 2003 (JSE)_ * These notes are organized by section numbers of the original document; not all sections are represented... ---++++ 1.3 Introduction.Motivation * Please change the "negatives" in this section (re: the current system) into "positives" about the planned system. i.e.: * There *will* be an explicit schema... * The output *will* be consistent with public schemas... * The use of RDF to model static/dynamic properties *will* be consistent with best practices... * The RDF models *will* facilitate validation against metadata schemas... * etc... ---++++ 2.1 Namespaces * The Dspace practice of "adopting" external schema without explicit reference to their external "homes" (and explicit relationships via the Harmony ontology) provides a stand-alone capability but seems like it would present long-term maintenance issues * Generally, everything implicit should be made explicit ---++++ 2.2 Resources * This whole section (and elsewhere whenever Harmony is referenced) could use a more formal/explicit expression of the relationship of Dspace-created or adopted properties to the Harmony ontology * I'm particularly concerned with the use of manifestation, item and bitstream * The references to Harmony need to be more *precise*; for example, nowhere in the ABC Model pubs can I find a definition of "bitstream" (Note: I think AndyS had similar issues) ---++++ 3.1.1 Use of External Schemas * No distinction should be made between semantically identical properties that are syntactically identical (same property name, same meaning) and those that are different (different property name, same meaning) * In other words, property relationships should be *explicitly* related across schemas by declaring (perhaps via subproperties) their shared ontology class * My concern that they are being treated differently is implied by the statement, _...properties in the local namespace that do not exist in...broader-purpose metadata sets but are related to them by meaning..._ ---++++ 3.1.2 Duplicate Properties * Same comment as above: there should be no duplicate properties, where relationships haven't been explicitly declared * Maintenance nightmare... * Is it possible to have a sameAs sub-property? * Meaning, to an application there appears to be two distinct properties * but they are in fact the *same* stored field (changing one changes the other) ---++++ 3.1.3 Resource Typing * Interpreting the syntax of a URL is a really bad idea over the long term * URLs should be treated as dumb numbers * resource types should be explicitly declared in metadata, period * This allows the use of other possible locator schemes over time, and facilitates migration, provisioning, etc. * Jason's suggestion (using the RDF-Schema type facility) should addresses this, but I don't see this from his example (using "rdf:Collection," whose type presumably has been defined, instead of "rdf:Description" ---++++ 3.2.1 Expression of Qualified Properties * Very good point about unqualified-DC clients ---++++ 3.2.3 Usage of Local URIs * I agree with Jason's point and recommendation ---++++ 3.2.4 Formatted text * Use of HTML would seem to present long-term and immediate alternative-usage issues why not use XML with appropriate XSL? ---++++ 3.3.2 Non-existent States et.al. * How are additions, deletions and migrations of properties (including adoption/application of namespace upgrades) encoded in history? Hope this is helpful...JSE | John S. Erickson, Ph.D. | Hewlett-Packard Labs | Norwich, Vermont USA | john.erickson@hp.com
Received on Wednesday, 7 May 2003 11:45:56 UTC