- From: Bill de hÓra <dehora@eircom.net>
- Date: Mon, 29 Apr 2002 10:13:11 +0100
- To: "'Joshua Allen'" <joshuaa@microsoft.com>, "'Danny Ayers'" <danny666@virgilio.it>, "'Sandro Hawke'" <sandro@w3.org>, <www-rdf-interest@w3.org>
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 > -----Original Message----- > From: Joshua Allen [mailto:joshuaa@microsoft.com] > Sent: 27 April 2002 18:55 > To: Bill de hÓra; Danny Ayers; Sandro Hawke; > www-rdf-interest@w3.org Subject: RE: Explicit Disambiguation Via > RDF bNodes, more Process > > > Let me first say that I am a bit uncomfortable trying to discuss > in terms of the "RDF Model Theory", since what I am talking > about is independent from RDF, and I tried to explain it in > plain English. But if I understand what you are saying, then I > completely agree. Bringing up the MT wasn't isn't a tactic to shift ground. Nonetheless there are restrictions on bnodes and it doesn't hurt to be middling precise. > I think that all of that formalism is simply saying "Whenever > something other than a URI is used to identify a thing, it is > impossible to know that two assertions are about the same > 'thing', even if the same words are used to identify the > 'thing'" That is exactly how things should be. I wasn't clear what 'x' was, I gather it’s a URI(ref) now. > In other words, if I said: > (http://www.microsoft.com/ isOwnedBy) hadIncome lots > and: > (http://www.microsoft.com/ isOwnedBy) residesIn prison > > The first (http://www.microsoft.com/ isOwnedBy) could mean > "Microsoft Corporation" and the second could mean "some guy who > owns one share of MSFT and is in prison". > > Clearly it would be invalid to assume that > (http://www.microsoft.com/ isOwnedBy) refers to the > same > entity in both assertions. > > Also note that the issue of whether or not I can merge two > different instances of (http://www.microsoft.com/ isOwnedBy) is > *not* the same as the way that I constructed the assertion in my > example. > > In my example, I was *asserting* that "The entity which owns > www.microsoft.com made lots of income" rather than *inferring* > it. That will work, but breaks down on data integration, which is a (the?) main ingredient of what we're supposed to be excited about. When we own all the assertions, we win, but that's a closed system. And I didn't need RDF to build that kind of system, I could have had CGs or KIF. The irony is that it's quite straightforward to merge RDF graphs syntactically. Digging around, Rob McCool and RV Guha are working on something that they think will alleviate the reference problem on an open system. It's an interesting idea, property centric, doesn't seem to require special scheme URIs. Possibly it makes an intersection between yours and Danny's positions, tho' I'm entirely sure of that. You can find it in the middle of this paper: <http://www.alpiri.org/sw002.html>, 4.1 Bill de hÓra -----BEGIN PGP SIGNATURE----- Version: PGP 7.0.4 iQA/AwUBPM0OouaWiFwg2CH4EQI86QCfbaL4hsFZYfXKDXJFxXua+i1Vt2YAnRzo xVU/J48alPf8lIxLFmuYRKSw =/dAE -----END PGP SIGNATURE-----
Received on Monday, 29 April 2002 05:20:11 UTC