RE: BioRDF: URI Best Practices

Matthias,

> On the page
> http://esw.w3.org/topic/HCLSIG_BioRDF_Subgroup/Tasks/URI_Best_
> Practices/Use_Cases
> I have written that an "ontology of resolvable resources" 
> would be practical in some cases.

If you read "the Architecture of World Wide Web
http://www.w3.org/TR/webarch/", you would know that W3C has already made the
distinction between the information resources and non-information resources.
(What you said the group one vs. group two).  

What you said is not correct that "When RDF was invented it was mostly
intended to be used with URLs of the second group", Tim Bernerds-Lee has
long envisioned that URI is for everything. A few days ago, Kerstin Forsberg
has posted a link http://www.ibiblio.org/hhalpin/irw2006/ about a recent
workshop.  The articles there will help you to get a grasp of the problem.  

Again, please don't mix naming issue with implementation/management issue.
The URI specification specifically said that we should not guess the nature
of the resource from the surface representation of the URI.  To figure out
what a URI represents, you must "follow your nose" to dereference the URI,
or other URIs ensued.

The first group is adequately solved with a HTTP 303 for the non-IR.  (the
http-range-14 issue, but see a description of it by Harry Halpin
http://www.ibiblio.org/hhalpin/irw2006/summary.html).

The distinction between what you called group 2 and group 3 can never be
solved with a just a name.  No one can promise that the metadata of a data
never change because in SW no one has the complete knowledge about a
resource.  RDF, and therefore SW, is based on open-world assumption. And
anyone can say anything about anything anywhere. What makes the metadata
about a resource is not definite in SW.  

To follow your example, let's talk about an image here.

1. http://www.charlestoncore.org/images/group-logo.png a
http://purl.org/dc/dcmitype/StillImage  .

There should not be any debate on if (1) is a metadata about the metioned
image.  But,

2. http://purl.org/dc/dcmitype/StillImage rdfs:subClassOf
http://purl.org/dc/dcmitype/Image .

First question, do you consider assertion (2) as a metadata about the image?
If so, you linked your metadata to an set of metadata (dublin core) that is
not under your control.  How can you assume their metadata, i.e., ontology
never changes?  And would you consider that 

3. http://www.w3.org/Icons/w3c_main a http://purl.org/dc/dcmitype/StillImage
. 

is a metadata about http://www.charlestoncore.org/images/group-logo.png ?

Second, each assertion itself does not have a URI, i.e., this assertions can
be placed in anywhere.  So, how can we obtain all the metadata about the
image? If we cannot, how can we assume there is no change in metadata.

We cannot talk bluntly about the metadata about a resource without a context
or closure.  And you cannot and should not infer the closure from the name.

 
With regard to the resolvable ontology, I am not sure if its intends.  Is it
to make claims about the URI or about the RESOURCE that the URI represents?
But in either case, the problem contradicts itself because: how how can
someone and somehow get a hold of a description in your proposed ontology at
the first place? There was a question in similar essense posted to the TAG
about if there should be a URI for the information resource or non-IR. I
don't think it gets any consideration from TAG either.

The distinction about the nature of the resource that a URI represents is a
very important one.  But just as Godel's incomplete therom has told us:
don't try to use the same technology to define the foundation of the
technology.  Doing so will only get more questions than answers.     

Cheers,

Xiaoshu

Received on Thursday, 20 July 2006 15:49:46 UTC