- From: Tim Berners-Lee <timbl@w3.org>
- Date: Fri, 3 Mar 2000 16:27:07 -0500
- To: "Pierre-Antoine CHAMPIN" <champin@bat710.univ-lyon1.fr>
- Cc: "ML RDF-interest" <www-rdf-interest@w3c.org>
I tim wrote: >> 0. We could define (if starting from zero) define dc:creator to have range >> dc:person where >> a person is the domain of properties dc:mailbox and sc:homepage and >> dc:commonname. >> That is the best solution. Pierre-Antoine replied: >do it, and people will write things like > ><play:Person rdf:about="mailto:John@somewhere.org"/> ><play:Person rdf:about="http://www.somewhere.org/~Paul/"> ><play:Person rdf:about="employee://somewhere.org/12345"/> > >et voila ! The above dc:creator statements are valid. They are valid in as much as we do not have any mechanism in RDF for causing a validity error. We can declare that "play:mailbox" has a domain of Person and a range of Mailbox, but we can't yet say that People and Mailboxes are mutually distinct. >This is what makes RDF flexible enough to scale the web. No, I don't think so. Vagueness is NOT what RDF needs. It is the ability to go in and annotate the relationship between two well-formed concepts from different places. For example, to say the rdfs:domain of play:mailbox is dc:Person is a useful link between schemata which can be written after play: and dc: have been defined by others. This is what allows RDF to scale as a well-defined web. >> >URIs are ambiguous, yes, they have more than one interpretation level, yes. >> >> No, no. We should be clean or no reasoning from all this will be possible. >> IMHO. > >My belief is that SOME reasoning will be possible, >even if hard logic inference (with completeness and everything) will not. It is entirely my intent that the platform will support hard logic inference, with completeness and everything. >We won't prove theorems, no Yes we will, this is entirely part of the idea. When you do ecommerce, you have to be able to *prove* that you have completed a transaction. > - but is there any universal truth among the web, anyway ? There is no universal truth in the Semantic Web except for mathematics. There will be sets of information which together form a consistent system. i expect also there will be some very large corpuses of consistent data such as the mass of catalogs and weather data and stock price information and so on. >> >We can't prevent people from using URI with different interpretations, Yes we can and we must. This is not going to be some sloppy HTML 2.0+- excercise. We can limit our procesing to to data we trust, and limit what we trust to data whose semantics is well defined. >> >so we'll have to use the context to tackle with it. >> >> we agree to differ then. > >well, I meant, we sure can ENCOURAGE people to use strict structures, >and if they do, RDF agents will be able to perform very efficient and powerful reasoning. >but we can't rely on it - if navigator did rely only on HTML recommendation, 3/4 of the web would be unreadable! For a reasoning engine, 99.999% is unreadable. That is whyc we need a semantci web which is different: hard logic. >one eroneous fact in a strict logical system can make the entire system contradictory ; A semantic web system always operates on a very small subset of the entire system. The SW engine will have to be always aware of where an RDF statement was found, whether in a document which is trusted (for what purpose?) or in a logical expression, for example. >the web can not afford strict logic : its first axiom is "there is contradiction". No, the web must have strict logic, and each agent on the web must be able in the end to exchange a proof of its conclusions based on classical logic -- even if it used soft heuristics to find that proof. >I'm thinking about writing a paper on the subject. >I'll post it on the list. I will be interested to read it. My thoughts are mostly in linked around http://wwww.w3.org/DesignIssues/Semantic Tim > Pierre-Antoine
Received on Friday, 3 March 2000 16:27:11 UTC