- From: David R. Karger <karger@theory.lcs.mit.edu>
- Date: Thu, 8 May 2003 14:07:00 -0400
- To: mick.bass@hp.com
- CC: mick.bass@hp.com, danbri@w3.org, mick.bass@hp.com, www-rdf-dspace@w3.org
some other comments on the document under discussion: Document lists properties of various (implicit) schema. Ultimately, we'll want an english description of each such property (in jason's document, and perhaps stored within the repository) Jasons suggests making assertions about relationship between eg dublincore properties and harmony properties. While it is not illegal for us to do this, I consider it a bit shaky: who are we to say what the harmony or dublincore schema writers think? we'll eventually have to deal with such translation issues (that is one of the key ideas of the semantic wb) but perhaps we sidestep for now. e.g, jason points out that hasPart appears in both dublincore and harmony. fine: let's use only one of the two hasPart properties. Having types is nice of course, but unclear to me why not having them "makes useful RDF queries more difficult to construct". we certainly should NOT assume type inference based on URL. General question: do we want to enforce schema validation within the dspace repository (big burden on the library) or do we just use schema as "hint"? Ie, where on the spectrum between "total schema compliance" (traditional database) and "do what you feel like" (haystackish) there are many operating points. although semantic web may _allow_ for sloppy data/schema violations, certain organization may be willing to commit to tighter obedience. is dspace one such? Didn't undertand section on "relationships based on local identifiers" d
Received on Thursday, 8 May 2003 14:03:21 UTC