Re: Correct Usage of rdfs:idDefinedBy in Vocabulary Specifications with a Hash-based URI Pattern

On Oct 21, 2010, at 10:05 AM, KangHao Lu (Kenny) wrote:

> Hello Martin,
> 
> I don't think my argument would be very logical, but we can't wait for rule engines to discuss this.
> 
>> Note, however, the majority of the Web vocabularies use the same URI for the entity name reference and the descriptor reference, see the link provided by Michael Hausenblas:
>> 
>> 	http://code.google.com/p/void-impl/issues/detail?id=45
>> 
>> and in particular the little survey by Richard Cyganiak posted on that page.
>> 
>> I personally would argue that in the case of ontologies / vocabularies, the conceptual difference between the entity and the descriptor is a lot less significant than when it comes to data, since an ontology is, by definition, a specification, i.e. a document.
>> 
> 
> Basically I like this approach, that is, I don't like the fact that some ontologies have '#' as end character and there should not a URI for an ontology document and a different URI for the *conceptual* ontology.
> 
> IIRC, 3 years ago Tim was very shocked by those ontologies that have '#' as end charter and claimed that this is not a good idea (and he would bring up this issue at TAG or awwaw, I can't remember). The argument was that string after '#' has the meaning of 'local identifier' (so that we use #I #i for WebIDs because 'I' is a 'local identifier') and identifiers can't be empty strings (or this might break some systems, I guess). I somehow agree with that, and Toby's use of "my:" to identify an Ontology makes me a little bit uncomfortable. I have no idea if there's any followup after Tim brought this to TAG or awwaw.
> 
> I have another argument, namely, you should distinguish the concept from the document only if the following criterion is satisfied.
> 
> - if the time when the thing with hash URI is created and the time when the document is created have *clear* difference
> 
> So this holds for people, so people should not use document URIs. This holds for organizations, cause you create the website of an organization maybe some years after the organization is founded. 
> 
> The problem is 'ontology'. I don't know whether you should call the structure an ontology or it became an ontology once it is written down, but I don't think the difference of the timing is very *clear*.

I agree its not as clear as the other cases, but an argument for making the distinction would be that the same ontology can be encoded by different documents. For example, an OWL/RDF ontology *is* an RDF graph, but that graph can be represented/encoded/choose your word ... in a variety of different documents with different syntax rules.  This is an old and familiar distinction, really, between type and token: one work called Moby Dick, many copies of it; one third letter of the alphabet, as many copies of the "C" character as you can shake a stick at, and so on.

> A similar example is when you want to give a URI to a python module. I would not end it with '#' because I don't see why we need do distinguish the 'module document' from 'module'.

That case is much blurrier, I agree. But imagine an algorithm implemented once in Python and elsewhere in C++. (Really, this is possible.) Same algorithm, very different documents. Ontologies are (arguably) more like algorithms than pieces of code.

Pat Hayes

> A module is a kind of document, so is ontology. So, owl:Ontology rdfs:subClassOf foaf:Document !
> 
> Well, this is a theory. If there's a common practice of using '#'-ending URI for ontologies, maybe we should accept it.
> 
> No strong opinion. Wasn't this discussed at AWWAW? Just curious.
> 
> Cheers,
> 
> --
> Kenny
> WebID: http://dig.csail.mit.edu/People/kennyluck#I
> What is WebID: http://esw.w3.org/WebID

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes