Re: URIs may not mean *exactly* what you think

(on-topic this time)

Charles McCathieNevile wrote:
> On Tue, 4 May 2004, Benja Fallenstein wrote:
> 
>>So it's not domain and range constraints that are the problem, it's
>>ignoring the URI owner's definition of a term, including those
>>constraints, that is the problem. Looking at the name of a property
>>doesn't suffice to use it correctly.
>>
> My point, at such length, is that there is no inherent reason why people
> cannot refine the definition of a property that someone else initially
> defined, and in some cases this is valuable. But as Benja points out, there
> are good reasons for being VERY conservative in doing so...
> 

I am not convinced that the concept of 'owner' of a URI is necessarily a 
good one, or appropriate for the semantic web. Judging by the web most 
uris have 'owner's who don't look after them much. Maybe it makes more 
sense for me to love and care for a URI that is already deployed, and 
liked by a community I work with, but is sadly neglected by its owner; 
rather than to create my own URI and give it a similar meaning. & even 
if I take this latter course, and then write some axioms linking the 
meaning of my URI with someone else's URI, I have just changed the 
meaning of their URI, just as much as I have changed the meaning of mine.

e.g. if I don't like dublin core I might choose to define my own 
property, which I may think of as a superproperty of dc:creator

e.g.

dc:creator rdfs:subPropertyOf my:agent .

and then say a comment

my:agent rdfs:comment
"This property links its subject with some agent that had
a causal influence on it, for example, authoring of or burning a book" .


===

my:agent may be a useful relationship; it is used in linguistics for 
instance.
It may be useful to show its relationship with dc:creator.
But my two axioms when combined with normal dc triples


eg:BasqueAuxiliaryVerbPaper dc:creator "Jeremy J. Carroll" .
eg:BasqueAuxiliaryVerbPaper dc:creator "Joseba Abaitua" .

(where the paper is the one mentioned in my earlier off-topic msg)
then suggest this agency relationship between strings and papers.

Even without this modelling clash, e.g. if we have


eg:BasqueAuxiliaryVerbPaper dc:creator _:jjc .
_:jjc vcard:fullName "Jeremy J. Carroll" .
eg:BasqueAuxiliaryVerbPaper dc:creator _:ja .
_:ja vcard:fullName "Joseba Abaitua" .

the dc:creator triple now means more (i.e. has more valid inferences) 
because of the definition of my:agent. Whether this additional meaning 
would be endorsed by the author of every dc:creator triple and the 
authors of Dublin Core, is unlikely.
===

So the concept of URI ownership in the semantic web is based on two 
fallacies:

1) you can distinguish a definition by the owner from a description by 
someone else

2) the meaning of one URI is independent of other URIs (possibly not yet 
coined)

(I suspect Peter could produce a few more fallacies)

Jeremy

Received on Thursday, 6 May 2004 03:04:55 UTC