RE: issues to be resolved before last call (rdfms-assertion)

> From: Dan Connolly [mailto:connolly@w3.org] 
> On Fri, 2003-01-17 at 08:15, Peter F. Patel-Schneider wrote:
> [...]
> > I would prefer it if the OWL documents could be silent on 
> this issue, and
> > let usage, convention, etc., determine social meaning, as 
> they have always
> > done.
> 
> The risks associated with doing that are unacceptably high, to me.
> 
> Namely, that people will claim that offers for sale, privacy
> and security policies, etc. *do not* have their usual
> social meaning (i.e. that their authors can be held accountable
> for thier contents) when written in OWL.
> This risk has already materialized for P3P.

Just a comment... they will claim that anyway, no matter what a
specification document says.  At best, the spec might be used in a court
of law to indicate what the originators of the spec might have thought
about the issue; however, in an English-style system of law where case
law is king, the spec cannot do anything much except inform the Court in
any particular case.  Only once a body of case law has built up can the
issue of social meaning be addressed in any legally meaningful way.

> For an elaborate rationale, see TimBL's essay...
> 
> "Bits mean something."
>  -- The Stack of Specifications
>  Tim Berners-Lee Date: 2002/05,
>  last change: $Date: 2003/01/06 19:40:09 $ 
>  Status: personal view only. Editing status: rough..
>  http://www.w3.org/DesignIssues/Stack

Yeeeesss... definitely worth a read.  See particularly the comment on
'Problem: this does not always happen' at the bottom of 'publishing an
XML document' and 'When the chain does not connect' - TimBL argues that
quoting is essential for bits to mean something.  Oh... guess what's
missing in everything except OWL Full... and we don't know how to do
interoperable things with Full as everyone will implement their own
(possibly incompatible) incomplete subsets for their own special
purposes such as calendaring.  And if the implementations are incomplete
in different ways, then how can we ever be sure that they will interpret
these extremely important bits in the same way?

I'm really not convinced that OWL, as currently specified, could be used
in the way detailed in that essay.

> >   However, the RDF Concepts document normatively states that the
> > entire social meaning of an RDF document is a part of the 
> RDF meaning of
> > that document.
> 
> Yes, that's as it should be. The RDF and OWL WGs are
> in the W3C Technology and Society domain for a reason:
> to see that the technologies are properly connected
> to the social systems around them.
> 
> >   Therefore I believe that the OWL documents must explicitly
> > disavow this view.
> 
> That's unacceptable to me.

Maybe OWL should pick up the view of 'social meaning' from the layer on
which it is transported.  In the case of OWL/RDF, it picks up the social
meaning part of RDF.  In the case of OWL/XML, it would have a formal
meaning but no social meaning.  In which case I'd use OWL/XML, despite
its non-normative status.  Requiring 'social meaning' and not having the
combination of acceptable quoting mechanisms and realistically
interoperable software would not be acceptable to me personally.

Note that I have not consulted Network Inference for company views on
this email; it is purely personal opinion.

		- Peter

Received on Friday, 17 January 2003 19:04:22 UTC