- From: Sergey Melnik <melnik@db.stanford.edu>
- Date: Thu, 14 Jun 2001 12:28:35 -0700
- To: Brian McBride <bwm@hplb.hpl.hp.com>
- CC: guha@alpiri.com, rdf core <w3c-rdfcore-wg@w3.org>
Brian McBride wrote: > > Sergey Melnik wrote: > > > With respect to this and other syntax issues, I'd like to remind of the > > "roundtrip" test, which have been raised many times on RDF Interest: an > > RDF tool must be able to parse, serialize, parse, serialize etc. without > > loss of information, i.e., on every parse, exactly the same set of > > statements is produced. Notice that after the first parse, rdf:ID will > > be necessarity replaced by rdf:about, since the model does not > > intrinsically capture the information about its origin. > > If > <rdf:Description rdf:ID="foo"/> > > is defined to be equivalent to, i.e. represents the same triples as: > > <rdf:Description rdf:about="#foo"/> > > and the round tripping test is defined to be that XML/RDF before > represents the same triples as the XML/RDF after (i.e. model > equivalence) which I think is the test that you suggest, > then this solution passes the round triping test. > > Brian This is right. However, by endorsing rdf:ID and rdf:about with relative anchors we embark on potentially unpleasant issues like e.g. HTTP redirect, support in RDF editors etc. In HTTP redirect, we'll need to specify precisely which URL will be assigned for the namespace, the original one, or the resolved one. Furthermore, if you are using an RDF editor, it has to maintain an additional flag with each "local" resource plus the URL of the page. This distinction becomes essential when you want to store the content at another URL and assert that it "moved from <previous URL>". If the editor cannot distinguish relative and absolute references, it will blindly replace <previous URL> with a relative reference resolving to the new page. Moreover, if the distinction between relative and absolute URLs is required, the parsers will need to support it as well. I'm trying to make a point that this issue is closely related to http://www.w3.org/2000/03/rdf-tracking/#rdf-ns-prefix-confusion, which has been resolved in favor of "in doubt be precise". Sergey
Received on Thursday, 14 June 2001 15:15:43 UTC