I really enjoyed the linked data session at WWW 2007:
- Use URIs as names for things
- Use HTTP URIs so that people can look up those names
- When someone looks up a URI, provide useful information
- Include links to other URIs so that they can discover more things
All Webby goodness! What’s not to like? But one issue has stuck in my head since asking the obvious question:
It seems that folks like Yahoo! and Twittervision offer different representations of the same resource using different URIs .xml, .json, .php, .yaml, etc. I’d say that’s arisen because users have little control over content-negotiation in their browser. Isn’t having the same resource identified by lots of different URIs an issue for the Semantic Web?
At the time Timbl offered the same answer as Danny, use OWL to say the two URIs are the same:
<http://DBpedia.org/resource/Berlin>
owl:sameAs
<http://sws.geonames.org/2950159>
But isn’t it be better just to use the Content-Location HTTP header in the reference? I’m almost ready to contemplate embracing RDF, but does that really mean I have to grok someone else’s out-of-band metacrap just to process it?
I’m also baffled why more RDF folks aren’t embracing OpenID, seems like every world changing demo I see still uses usernames to identify users, not URIs. What’s that all about?
Technorati Tags: HTTP, openid, semanticweb
In the RDF world you can have URIs for cities, and other URIs for documents *about* cities. The distinction matters. In the REST world you can only have URIs for documents about cities. REST folks always gloss over this.
Content negotiation applies only to document URIs, not to the other kind of “pure resource identifier”. The DBpedia and Geonames URI from your example are not URIs for a document about Berlin; they are URIs for the city itself. HTTP is not equipped to assert the equality of non-document resources.
Yeah, but given the .yaml URI it would be nice to know that it was a representation of “Berlin”. What I like about linked data is it puts the Web back into the Semantic Web. Just minting URIs without a representation ever existing is similar to Web Services and XML namespaces, something we’ll look back on and regret.
“HTTP is not equipped to assert the equality of non-document resources.”
Richard, you in turn glossed over the information resource compromise aka http://www.w3.org/2001/tag/issues.html#httpRange-14. Norm has by far the clearest explanation of what is a very scary compromise, the meaning depends on the server’s response code:
http://norman.walsh.name/2005/06/19/httpRange-14
That way lies madness. I hope the semweb community aren’t going to implement that just because denotation and description are a sticking point for the model theories.
I don’t find httpRange-14 particularily scary. The compromise has ended years of uncertainty about how to use RDF on the Web; It is no coincidence that Linked Data has sprung up shortly after the compromise.
“..seems like every world changing demo I see still uses usernames to identify users, not URIs. What’s that all about?”
___
You obviously weren’t really paying attention in /my/ WWW2007 talk (you were just taking blurry pictures at the back, I see)! =0)
User microresources (with URIs!) ‘hydrated’ in, and animated by, the browser were pretty core to my (world changing) Micro Web demo…
Let’s get together in a pub so I can have another go at explaining it all to you. =0)