W3C home > Mailing lists > Public > www-rdf-dspace@w3.org > December 2002

Re: FW: Fine-grained representations

From: Mick Bass <bass@MIT.EDU>
Date: Tue, 10 Dec 2002 23:40:36 -0500
Message-Id: <4.3.2.7.2.20021210223448.020888c0@po10.mit.edu>
To: "David Francois Huynh" <dfhuynh@ai.mit.edu>, <haystack@theory.lcs.mit.edu>, www-rdf-dspace@w3.org
reply crossposted to www-rdf-dspace

David Huynh wrote:
 > In the course of exchanging information with the outside
 > world, annotations are sometimes removed from information
 > (intentionally or unavoidably). Many examples come to mind:
 > 1. Converting the Xenon content of an e-mail message (and
 > many of its fields) into plain text.
 > 2. Converting the Xenon content of a document into HTML.
 > 3. Taking a snapshot of the UI in the form of HTML (for thin client).
 > 4. Exporting BibTex entries (not all annotations are lost).
 > 5. Printing.
 > 6. Speech output.


 > (Note that image is a media type, but it can carry a lot of
 > different "user types": x-ray prints, photographs, maps,
 > engineering blue prints, sketches, paintings, etc. Once
 > everything is translated into the digital world, what's going
 > to differentiate an x-ray print from a photograph?! More on
 > this later.)

 > Sometimes, a navigation to an object should not bring up just
 > one view. Consider this popular line from most detective
 > action movies: "So what do we know about this guy John Doe?"
 > The results of such a query would include photographs,
 > fingerprints, crime records, etc. That is, many projections
 > of the same entity onto many different domains. Perhaps our
 > view navigator should also show many views for certain objects?

--

The questions that you ask are also relevant for DSpace and SIMILE.

--

There is much prior thinking in this area.  In particular the OAIS 
reference model may be useful.  While it originated in the space data 
community in the context of long-term archiving of datasets, it has 
something to say about the questions that you ask, especially in the 
relationship between (in their terminology) "Archival Information Packages" 
(AIPs) and "Dissemination Information Packages" (DIPs).  DIPs are basically 
contextually-relevant views on an AIP.  The AIP of course then must contain 
a superset of information, sufficient to generate any requestable DIP.

The reference model is a bit of a verbose read, especially at first sitting 
(although by second or third sitting one finds oneself warming up to it 
like a good novel).

OAIS is described in full at 
<http://ssdoo.gsfc.nasa.gov/nost/isoas/ref_model.html>.

Brief overviews are available at:
- 8-page overview, 
<http://www.nasda.go.jp/pr/event/app/spaceops/paper98/track3/3b002.pdf>
- OCLC newsletter pages 26-30, <http://www2.oclc.org/oclc/pdf/news243.pdf>
- html slide set 
<http://ssdoo.gsfc.nasa.gov/nost/isoas/presentations/USDA_19990219/sld015.htm>

--

Some architectures for managing the transform of  AIPs to DIPs by binding 
objects to transforming services have been proposed and 
implemented.  Haystack is one.  Another is FEDORA, which you should be 
aware of as you dig further into answering the questions that you pose.

- http://fedora.comm.nsdlib.org/techdoc.shtml

--

Regards,

- Mick

=============================================
Mick Bass, Sloan MOT 2000

R&D Project Manager, Hewlett-Packard Company
Building 10-500 MIT, 77 Massachusetts Avenue
Cambridge, MA 02139-4307

617.253.6617 office    617.452.3000 fax
617.899.3938 mobile    617.627.9694 residence
bass@alum.mit.edu      mick_bass@hp.com
=============================================
Received on Tuesday, 10 December 2002 23:40:49 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 13:35:19 EDT