RE: Using URIs to identify non-information resources

Jacco,  Thanks for the feedback and questions!  Answers and comments are
inline below.

> From: Jacco van Ossenbruggen [mailto:Jacco.van.Ossenbruggen@cwi.nl]
> . . .
> A standard way to generate a URI for the former given the URI of the
> latter feels as a useful part, but still, only a part of complete 
> solution that can be worked into a best practice guideline. 

Agreed.

> Being able
> to specify the relation between the two URIs is another (again, the
> rdf:definedBy approach feels as only part of the solution).

Yes, from reading the RDFS specification, it looked to me like
rdfs:isDefinedBy[4] would be an appropriate predicate to express this
relationship, such as:

	http://thing-described-by.org?http://dbooth.org/2005/dbooth/
		rdfs:isDefinedBy 
			http://dbooth.org/2005/dbooth/ .

However, I don't know if this is how people are using this predicate in
practice, which is why I posted separately[5] asking about its usage.

In any case, although I think best practice should be to provide this
information, I view it as being independent of the thing-defined-by.org
suggestion.

> Questions I was asking myself about your thing-described-by.org 
> proposal include:
> - If you want to change the URI of your homepage, do you also want to
> change the URI of yourself?  

Maybe, but that's an uncool[6] thing to do if it causes users of your
old URIs to experience broken links.  :)

> If not, the dereferencing gives you the
> "wrong" document. So what do you do?

Either provide forwarding information or do a better job of choosing the
URI from the beginning, so that you won't have to move it.

No matter what solution you choose, if you offer the friendliness of a
URI that is browsable, then you incur the risk of changing your mind
about where the describing document should be.  This is mostly
independent of thing-described-by.org, since the document URI that you
provide is entirely your choice.  

If Dan Brickley's TAG proposal to sanction a 302 response[8] is accepted
as an alternative to 303, then you could use purl.org URIs instead of
thing-described-by.org URIs.  I think the pros and cons of purl.org URIs
for this use would be the same as for other uses.

Pros of purl.org:
	- you can change the document URI later

Cons of purl.org:
	- document URI must be pre-registered with purl.org.  
	- requires an extra HTTP request to determine the document URI

> - What if you use an HTML home page as the target URI in the 
> thing-described-by.org (as you do now) and want to change that to an 
> RDF file (e.g. at some point in the future you want to use your 
> RDF/FOAF profile instead)?
> - what if you want both the HTML and RDF option?

A few approaches come to mind:
	- Use HTTP content negotiation to serve the desired format
	- Use GRDDL[9]
	- Use RDF/A[10]


> - what if you do not control the URI of the physical thing?

The thing-described-by.org proposal is only relevant when you are the
owner of the URI of the physical thing and you wish to associate an
authoritative document with it.  (By "authoritative", I mean that you
are the owner[7] of the URI of the physical thing, and thus have social
license to define its meaning.)

> - etc...
> 
> Or more general: it feels a bit like the "wrong" way to do
> things: you 
> define something "important"(the URI for yourself, a thing in the 
> physical world) in terms of something "less important"  (some digital 
> artifact or information resource that is likely to change, 
> even if you 
> would use things like purl.org etc)[1]. 

Bear in mind that the URI for the important thing can be entirely
arbitrary, provided that:
	a) it is globally unambiguous; and
	b) it is potentially browsable to useful, authoritative
information.
So I don't think there's anything inherently wrong about basing the URI
of the important thing on the URI of a describing document.  That's no
more arbitrary than basing it on a random number or sequentially
allocated number.

> In connection to the multimedia task force: A similar problem
> occurs in multimedia annotation. Since the multimedia resource 
> being described almost always represents something in the 
. physical world, one wants to 
> distinguish meta data about the multimedia resource (e.g. 
> #rembrandtPainting dc:format 'image/jpeg') from meta data about the 
> physical work (#rembrandtPainting dc:format 'oil on canvas'). 
> . . .
> Comments welcome (and David, I'm sorry if this is a
> distraction from the discussion you want to provoke)

That is an interesting problem, but yes, I do view it as independent of
the thing-described-by.org proposal, which is really predicated on the
assumption that you *do* want to have separate URIs for the physical
thing and a document describing that physical thing, but you want the
URI of the physical thing to be browsable to a useful document
describing it.

> 
> Jacco
> 
> [1] Could be me just being old fashioned: I just need to
> understand that all
>      the real things around me are defined by their digital 
> counterparts, but I'm not there yet :-)
> [2] http://www.vraweb.org/vracore3.htm
> [3] not sure how to translate the concept of a Dublin Core or 
> VRA Core 
> "record" to RDF... Reification might not be the best way to do it.  
> Maybe named graphs are an option ...

Other references:
4. rdfs:isDefinedBy: http://www.w3.org/TR/rdf-schema/#ch_isdefinedby
5. Usage question:
http://lists.w3.org/Archives/Public/public-swbp-wg/2005Jul/0061.html
6. Cool URIs Don't Change: http://www.w3.org/Provider/Style/URI.html
7. URI ownership: http://www.w3.org/TR/webarch/#uri-assignment
8. TAG 302 question:
http://lists.w3.org/Archives/Public/public-swbp-wg/2005Jul/0056.html
9. GRDDL: http://www.w3.org/TR/grddl/
10. RDF/A: http://www.w3.org/MarkUp/2004/rdf-a.html 

David Booth

Received on Wednesday, 27 July 2005 19:22:37 UTC