On 11/10/10 3:59 PM, Lars Heuer wrote:
Hi Kingsley,
Thanks for your reply.
[GET<http://en.wikipedia.org/wiki/John_Lennon>]
That's a Document Address, by default i.e., HTTP 200 OK response when
you HTTP GET.
ACK.
Let's assume Wikipedia would return 303 like DBpedia does. Does it
solve the problem?
No, they would have to implement a disambiguation heuristic using 303
that separates Document Address from Entity Name, assuming they adopt
what is known as a slash terminated style of URI re. Linked Data.
Why? Doesn't the response depend on the requested content/media type?
If I want an RDF/XML representation of the document, I can ask for
Accept: application/rdf+xml
and Wikipedia would (ideally) return an RDF/XML representation of that
resource which tells me that John Lennon is a person who was born at
... murdered at ... was part of a group named ... etc.
Yes, so you received a document stating all of the above, who is the
Subject? How is the Subject Identified?
Have to drop the fact that your non-web-sign-processor (DNA CPU)
already groks "John Lennon", and does a lot of fancy processing with
frames en route to disambiguation and context manifestation.
Web isn't anywhere close to the Human Brain (not that I have 100%
comprehension of how it works, but from the little I understand, it
trumps anything produced by computers so far).
I think, it does not solve it, since I cannot make
statements about the *page*
<http://en.wikipedia.org/wiki/John_Lennon> (since I always get 303 and
an agent would interpret it as NIR).
If they adopt the heuristic in play re. DBpedia, it will be fine.
1. http://dbpedia.org/resource/John_Lennon -- Name
2. http://dbpedia.org/page/John_Lennon -- HTML Document with RDFa inside
I see, DBpedia provides different IRIs. That's fine. But it's not
possible to keep<http://en.wikipedia.org/wiki/John_Lennon> (or
<http://dbpedia.org/resource/John_Lennon> if that matters) and make
statements about that, right? I cannot make statements which are
interpreted rightly without an Internet connection. I need the status
codes.
[...]
Personally, it can be solved at the application level by application
developers making a decision about the source of semantic fidelity i.e
HTTP or the Data itself.
To take an practical example: If I want to make statements about the
following NIR:
<http://psi.connectors.de/product/8974>
What would I have to do? Do I need a redirect? Why? If the above
mentioned IRI would return RDF/XML (or any other media type requested
by the client), why do I need a 303? 200 + requested media type +
content should be enough, shouldn't it?
I am not a proponent of one size fits all re. heuristics for resolvable
Identifiers re. Linked Data.
If you are developing a Linked Data app. and you make a commitment in
your code to self-describing data, then 200 OK + Content-Location header
+ Structured Document (a Descriptor) is fine. Your data handles the Name
vs. Address disambiguation.
[...]
Ian is indicating that RDF based Linked Data should dog-food i.e., if
RDF formats are about the content of structured data documents, where
the data describes itself, who is HTTP to determine otherwise re. Name
or Address? :-)
I am unsure if I am falling into the same trap?!?
Why is it a trap?
What Ian proposes is an option.
Side note: Each subject/object needs a GET (assuming that predicates
are always NIRs) to interpret the statement correctly... Does it
scale? Let's assume you'd send me a DBpedia dump. I cannot interpret
it correctly, unless I have an Internet connection?
What about when I send you DBpedia in the post on a USB key ? :-)
Best regards,
Lars
--
Regards,
Kingsley Idehen
President& CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen
