Re: httpRange-14

* Tim Berners-Lee <timbl@w3.org> [2003-07-28 11:28-0400]
> 
> On Saturday, Jul 26, 2003, at 12:12 US/Eastern, Dan Brickley wrote:
> 
> >* Tim Berners-Lee <timbl@w3.org> [2003-07-25 20:51-0400]
> >>
> >>
> >>The nearest cyc term is (from previous discussions)
> >>cyc:ConceptualWork.  This something abstract, not concrete.
> >>
> >>The word "documents" has been used on this thread because it is 
> >>shorter.
> >
> >>From previous discussion, am I right in saying you'd count
> >"The Bible", "Hamlet" etc as conceptual works; but foaf:Document (the
> >RDF class) or cyc:ConceptualWork (another RDF class) as not being
> >conceptual works.
> >
> >If so, what is it about the former that makes them more 'conceptual'?
> 
> 1. I point you to a whole ontology - check the cyc classifications.
> 
> 2. Nothing.  They are all conceptual. Only things which convey 
> information
> are works, in the sense of opus, oeuvre, something which decreases 
> entropy
> and hence takes energy to create.

There's the basis of a distinction there, although the connection with 
the HTTP URI scheme naming policies is pretty weak.

> >Are you claiming that RDF classes are purely mathematic constructs,
> >discovered rather than created?
> 
> I make no untestable statements about whether they are discovered or 
> created.
> But I do distinguish them from for example an RDF schema, which has 
> content
> and can be represented in RDF/XML or N3, can be copyright, etc etc.
> 
> >I ask because I have several RDF implementations which depend on 
> >classes
> >being allowed in httpRange...
> 
> Could you manage a transition to one in which you use  a hash, please?
> It is going to be easier to fix it now than later.

For FOAF, we could probably pull the switcheroo now with a fair amount
of annoyance for implementors. But FOAF files also use Dublin Core,
RSS1, MusicBrainz, XMP etc namespaces, all of which are named with a
'/'. Also they use my wordnet namespace...

The wordnet case is trickier. We have a namespace with 50,000 terms
(classes) in it, named like http://xmlns.com/wordnet/1.6/Tree and 
dereference-able. There is no obvious parallel design with a # that
allows both (a) reasonably intuitive end-user markup in the instance
data and (b) terms to be individually derferenceable to something
useful and smaller than the entire multi-megabyte dictionary. 

Do you really think it's worth pulling the rug out from under the feet
of 80%+ of RDF early adopter vocabularies? Just as things are starting
to take off... Seems risky to me, for a somewhat academic point... If we
do it we'll need a damn good explanation for the disruption.

Dan

Received on Monday, 28 July 2003 11:43:05 UTC