Social meaning undermines semantic web
What is the point of the ``social meaning'' stuff? Is it supposed to indicate that RDF documents available on the web are not always supposed to be considered to be assertions? If so, how is this done? Can I, for example, use rdfs:comment to put disclaimers into an RDF document? If there is no way in RDF to make such disclaimers, then why bother to bring up the possibility? I find the whole example about clowns to be completely mystifying. If you take this example at face value, then *any* use of any RDF commits to the natural language implications of rdfs:comment tags. How can any organization deploy an RDF-aware application under these circumstances (except by having that application understand the implications of arbitrary natural language). Similarly, the tying of the meaning of a URI to the ill-specified intent of some organization poses a giant bar to the deployment of RDF. Under these circumstances how can any organization use an URI that they do not own? The owning organization might, after all, choose to change the meaning of any URI they own at any time. This seems to me to be a bar to any communication between organizations using RDF.
See: http://lists.w3.org/Archives/Public/www-rdf-comments/2002OctDec/0053.html
See: 024-Various.html
My precis: see original for detail.
What is the point of material on social meaning? If not all RDF on the web is assertions, how is this indicated, or if there is no such way why mention it? Committing to the meaning of natural language tags: how can any application do this without natural language understanding? Committing to the meaning of URI as "ill specified" intent of some organization, which may be changed at any time. How can anyone commit to anything involving a URI they don't own?GK to field issue.
The point of this material is that to be useful, RDF has to operate in a social context. Otherwise it's just an academic exercise.
The mechanism for indicating assertion vs non-assertion is not defined here. RDF is part of a larger system. Concerning the need to understand natural language: I'm having problems understanding the concern here. I don't recognize the problem raised. There's no requirement for applications to understand natural language. I should think about an e-commerce example to illustrate this. Committing to intent: I suppose that in a high-value environment, one would have to be very careful whose URIs you used. Again, I think there could be an illuminating e-commerce example here. On commitment: note that it is not an application that makes a commitment by publishing RDF, but a person or organization who causes the material to be published. So committing to a statement that uses someone else's vocabulary involves an element of trust, like almost any social interaction.Consider an e-commerce application: a quotation and corresponding purchase order may be rendered as RDF, thus creating a legal contract. The quotation contains a human-readable description of goods to be supplied. In normal operation, an invoice (also rendered as RDF) would create a legal liability on the customer to make the payment indicated, provided that the goods described have indeed been supplied. This liability does not depend on any software's ability to understand the human-readable description of the goods.
In a plausible implementation, the software systems operated by the customer would require an additional input in the form of a confirmation from a suitably authorized person that the goods had been received before responding to an invoice with a release of payment. Although related, this is a distinct from the issue of liability addressed in the document.See: http://www.ninebynine.org/wip/RDF-concepts/2002-??-??/TBD
The text concerned has been revised and simplified for the next WS rekease. But the fundamental principle remains. If this continues to be a point of concern then it should be subjected to wider discussion during last call.