RE: Issue rdfms-abouteach

> -----Original Message-----
> From: ext Dan Brickley [mailto:danbri@w3.org]
> Sent: 16 November, 2001 14:38
> To: Stickler Patrick (NRC/Tampere)
> Cc: dave.beckett@bristol.ac.uk; w3c-rdfcore-wg@w3.org
> Subject: RE: Issue rdfms-abouteach 
> 
> 
> On Fri, 16 Nov 2001 Patrick.Stickler@nokia.com wrote:
> 
> >
> > > For writing by hand, there are, ahem, non-XML syntaxes that
> > > are better.
> >
> > I'm not saying there aren't. But the spec defines an
> > XML serialization and that is what the masses will use.
> >
> > IMO, anyone who things that the web community at large will
> > adopt a non-XML notation for the interchange of RDF encoded
> > knowledge is mistaken.
> 
> Heres a spin on things: this isn't just about interchange. 
> The ability to
> say about-each-ish things like 'all W3C employees have an interest in
> technology'; 'all W3CRecommendations have a dc:format of 
> text/html' etc is
> useful both within RDF/XML documents that are exchanged, but 
> also within
> databases of RDF content. The aboutEach mechanism is useless for that
> purpose, though this may not be clear to content producers.
> 
> We know from the past 3-4 years that content producers often have
> difficulty thinking about RDF separately from its XML syntax. The
> aboutEach mechanism is indeed a handy way of writing down in 
> markup some
> simple rules about common properties that apply to collections of
> resources. If you know what you're doing. The real problem here isn't
> about parsers, streaming etc., but about the risk of encouraging RDF
> content producers to statically encode such rules in RDF/XML 1.0's
> aboutEach syntax. By doing so, they are writing down
> useful knowledge in a form that will (a) be lost forever when 
> the markup
> is parsed (b) be difficult to convert into more explicit rule 
> and ontology
> formats that can live in databases, behind APIs etc., since 
> the authorial
> intent behind these syntactic shortcuts will be invisible to automatic
> tools.
> 
> I for one will never enourage people to write down useful 
> generalisations
> in aboutEach syntax, because I don't want to have them come 
> back and ask
> me why those rules aren't accessible via the (graph-oriented) 
> APIs, query
> languages, database interfaces etc that they'll have to use to access
> their content. In my experience of talking to RDF developers 
> _and_ content
> producers, there's often misunderstanding about which 
> features of the XML
> syntax are carried through to the abstract graph. So my problem with
> encouraging the use of aboutEach is that it risks creating a 
> huge legacy
> problem: information loss as we go from the RDF/XML into 
> databases, APIs
> etc. Because about aboutEach mechanism _appears_ to be RDF's 
> way of making
> generalised claims about members of a collection, people will 
> likely use
> it as such unless we attach a health warning. Once it becomes 
> clear that
> aboutEach is just a wierd macro mechanism, I believe it'll 
> lose its appeal
> to content producers.
> 
> 
> Dan

Point taken. And you are likely right about that.

I'm not 100% convinced that aboutEach should be dropped now, until
a *better* way to do things is available, but I am sufficiently
swayed by yours and others arguments to defer it to the rest of
the WG and abstain from expressing further opinion.

Regards,

Patrick

Received on Friday, 16 November 2001 08:11:46 UTC