[pedantic-web] Linked Data Service of German National Library - URI issue

[Neubert Joachim]

Last week, the German National Library (DNB) has published the authority
data of the German libraries on the web. That's really great - the data
covers 1.8 million persons, 1.3 million organizations and 170,000
subject descriptors, all of it carefully crafted. The RDF/XML dataset
can be downloaded and is, according to the project page
(https://wiki.d-nb.de/display/LDS, in German) free for non-commerical
use.

However, the Beta version of DNB's linked data service has an URI issue
which may be quite severe for its use in linked data environments:

The German writer Bertold Brecht is found at

http://d-nb.info/gnd/118514768

This URI gives a HTML document with lots of information and links about
Brecht, if HTML is requested (200, without any redirect). If RDF/XML is
requested, it redirects per 302 to another URI,

http://d-nb.info/gnd/118514768/about

which delivers RDF/XML.

For preceeding prototypes, several people had pointed out that it's a
bad idea to use the same URI for the person and for the HTML document
about the person, that this contradicts Linked Data Principles, and how
URIs could be built in a "cooler" way. Nonetheless, the new Beta version
of the service (which claims a "preliminary stable state") uses the same
URI scheme.

In the FAQ the idea behind this is explained
(http://www.d-nb.de/hilfe/service/linked_data_faq.htm#uri). The authors
admit that it may be desireable to have separate URIs for a concept and
its description. However, they add, in this case a separate URI for the
HTML representation should not be necessary, because they themselves
nowhere refer to the URI of the HTML representation. They offer to adapt
this at a later date if URIs to the HTML representation are requested by
users.

I think it could be very helpful if people on this list could point out
what exactly - which tools and which processes in the web of data - are
broken by such an approach, and how it could be improved in order to add
the full value of this great data source to the Linked Data Web.

Cheers, Joachim

[Dan Brickley]

On Mon, May 10, 2010 at 1:50 PM, Neubert Joachim <J.Ne...@zbw.eu> wrote:
> Last week, the German National Library (DNB) has published the authority
> data of the German libraries on the web.

[...]

> I think it could be very helpful if people on this list could point out
> what exactly - which tools and which processes in the web of data - are
> broken by such an approach, and how it could be improved in order to add
> the full value of this great data source to the Linked Data Web.

Just as an aside --- and no criticism intended re Joachim's nice post ---

While I'm sure constructive criticism would be useful and appreciated,
I hope everyone remembers to say nice stuff about these linked data
announcements too. As we get more and more linked RDF in the Web, it's
easy to forget how cool each new dataset really is, and to say thanks
to the folk who worked on it. So for example the announcement at
http://lists.w3.org/Archives/Public/public-lod/2010Apr/0321.html
didn't appear to get a lot of commentary. Maybe we're getting too used
to success? ;)

cheers,

Dan

[Richard Cyganiak]

On 10 May 2010, at 12:50, Neubert Joachim wrote:
> I think it could be very helpful if people on this list could point
> out
> what exactly - which tools and which processes in the web of data -
> are
> broken by such an approach,

Under what license is the data published? Who owns the copyright?
Couldn't they make this explicit using dc:license and dc:rights? Do
license and copyright apply just to the RDF, or also to the HTML?

I note that parsing the HTML with an RDFa parser yields the following
assertions:

<http://d-nb.info/gnd/118514768>
html:chapter <http://www.d-nb.de/aktuell/index.htm> ,
<http://www.d-nb.de/netzpub/index.htm> ,
<http://www.d-nb.de/sammlungen/index.htm> ,
<http://www.d-nb.de/service/index.htm> ,
<http://www.d-nb.de/standardisierung/index.htm> ,
<http://www.d-nb.de/wir/index.htm> ;
html:start <http://www.d-nb.de/index.htm> ;
html:icon <http://d-nb.info/favicon.ico> ;
html:stylesheet <http://d-nb.info/css/print.css> ,
<http://d-nb.info/gnd/js/jscalendar/calendar-win2k-cold-1.css> .

So they are already making assertions about the HTML.

> and how it could be improved in order to add
> the full value of this great data source to the Linked Data Web.

Hash URIs.

<http://d-nb.info/gnd/118514768> is the generic document.
<http://d-nb.info/gnd/118514768#person> is the person described therein.

Easy to implement and conceptually straightforward. What's not to like?

Best,
Richard



>
> Cheers, Joachim
>