- From: <noah_mendelsohn@us.ibm.com>
- Date: Tue, 24 Sep 2002 11:55:56 -0400
- To: "Jonathan Borden" <jonathan@openhealth.org>
- Cc: bwm@hplb.hpl.hp.com, skw@hplb.hpl.hp.com, tbray@textuality.com, timbl@w3.org, www-tag@w3.org
With apologies in advance for commenting on a narrow point in a thread that I probably don't completely grok in the larger sense: Experience with compound document systems suggests that we should indeed "go slow" in allowing embedded RDF to be interpreted independent of the semantics of its container. My favorite compound document example is a fictitious word processor that chooses to keep it's "undo" list persistently in the document along with the current text. Thus, assuming wp is the word processor's namespace, we get something like: <wp:document> <wp:currentContent> ...real stuff here.. <rdf:xxx ...probably describes the real stuff../> </wp:currentContent> <wp:deletedStuff> ...stuff that used to be in the document.. <rdf:xxx ...probably describes the stuff that's no longer in the document../> </wp:deletedStuff> </wp:document> Let's say that my document used to say that the author was Noah, and the editing tool was smart enough to capture that in an RDF assertion. Now I change my mind and delete that statement of authorship from the document. The RDF (probably) goes to the undo list, so that it can reappear if I issue an "undo". It would clearly be a mistake to trust that RDF as representative of the current state of the document. Of course, you could require that the application either edit the RDF to assert "we used to think Noah was the author but we don't think so now", or to use a non-RDF representation. I think that, in general, such approaches reduce the power of XML. Let's say I use XML to dump a database that contains XML fragments in various records. Further assume some of that XML contains RDF assertions. Should it be the responsibility of the dump program to rewrite those assertions to say: these assertions don't relate to the dump format container, they relate to the database records from which the dump was taken? I don't think so. All of the successful hierarchical compound document systems I know give outer structures relatively complete control over the semantics of anything contained within. I suggest great caution if any other approach is tried with embedded RDF. Again, apologies if my lack of understanding of RDF nuances has somehow reduces the usefulness of this comment. ------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 IBM Corporation Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------
Received on Tuesday, 24 September 2002 11:59:39 UTC