[whatwg] RDFa Problem Statement

Kristof Zelechovski wrote:
> Web browsers are (hopefully) designed so that they run in every culture.  If
> you define a custom vocabulary without considering its ability to describe
> phenomena of other cultures and try to impose it worldwide, you do more harm
> than good to the representatives of those cultures.  And considering it
> properly does require much time and effort; I do not think you can have that
> off the shelf without actually listening to them.
> In a way, complaining that the Microformats protocol impedes innovation is
> like saying 'we are big and rich and strong, so either you accommodate or
> you do not exist'.  Not that I do not understand; it is straightforward to
> say so and it happens all the time.
> Chris

Let me give a quick example of how this works in RDFland.

Each vocabulary defines nothing except classes (types of thing) and 
properties (aka relationship types). In FOAF for example, we defined 
Person, Agent, Document, OnlineAccount, Project, Group as classes. And 
we defined properties too. These tend to have a bit more 'character' 
than the classes, and carry the distinctive style of each vocabulary. 
FOAF has properties of Person and Agent such as 'openid', 'homepage', 
'weblog' that have as their range (ie. values) instances of the class 
Document. We also define properties like 'primaryTopic' that relate a 
page primarily about something to the thing itself. Each class and 
property is considered to be in the vocabulary whose URI is 
http://xmlns.com/foaf/0.1/ ... and this is the basis of RDF's "division 
of labour" mechanism. See also a squiggly diagram at 
http://danbri.org/2008/foafspec/foafspec.jpg (apologies that this is 
currently inaccessible).

The SIOC project declares a bunch more classes and properties. Some of 
these are defined with relationship to Person, Document, OnlineAccount 
from FOAF; classes that sub-class ours, or properties that cite our FOAF 
classes as the range or domain. DOAP does the same, expanding from the 
class Project to describe opensource projects. I've talked about this 
before so won't go on about those schemas.

The point about cultural diversity, independent extension etc is made 
better by the JaUranai FOAF extension that appeared a few years back:

http://kota.s12.xrea.com/vocab/uranai

They decided that FOAF was nice and all but was lacking some properties 
important in a Japanese context. So they declare new RDF properties: 
starsign, bloodtype, and various others that I don't fully understand 
because they have japanese names and documentation. From blood type's 
description from the RDF Schema file at 
http://kota.s12.xrea.com/vocab/uranai/uranai.rdf

<rdf:Property rdf:about="http://kota.s12.xrea.com/vocab/uranaibloodtype">
  <rdfs:label>???</rdfs:label>
  <rdfs:label xml:lang="en">Blood type</rdfs:label>
  <rdfs:comment>?????????</rdfs:comment>
  <rdfs:comment xml:lang="en">A blood type.</rdfs:comment>
  <rdfs:domain rdf:resource="http://xmlns.com/foaf/0.1/Person"/>
  <rdfs:range rdf:resource="http://www.w3.org/2000/01/rdf-schema#Literal"/>
[...]
</rdf:Property>

This effectively wires in 'bloodtype' to the other classes in use in 
this wider community. Wherever SIOC or DOAP projects have created a 
property whose range is "Person", we know that Uranai's 'bloodtype' 
property is also applicable. Without needing heavy duty coordination 
between the SIOC and DOAP authors and the author of Uranai.

Furthermore, the fact that all these projects share a common syntactic 
grammar means that I can simply add a Uranai 'bloodtype' property into 
my FOAF self-description, and expect each and every RDF parser and 
SPARQL database to immediately be able to parse and query it - see 
http://danbri.org/words/2008/02/25/286 for example. As Manu describes in 
http://blog.digitalbazaar.com/2008/08/23/html5-rdfa-and-microformats/ 
this is rather different to the Microformats.org approach, which is by 
intention a monolithic community designing a single, self-consistent 
product.

Back on my point that RDF vocabulary classes (ie. named types of thing, 
Person etc) tend to be boring, and the properties more interesting. This 
is to address the difficulty you mention, ie. ... "If you define a 
custom vocabulary without considering its ability to describe phenomena 
of other cultures and try to impose it worldwide, you do more harm than 
good to the representatives of those cultures".

So for example in FOAF, we define fairly boring bland classes (like 
Person, Document) in a way that allow different cultures to attach 
properties that they care about. It seems "bloodtype" is more important 
in Japanese culture than in Western Europe, but that the toolset and 
design provided by RDFa allows independent extension of FOAF in Japan 
without expensive central bottlenecks. For Creative Commons, they have 
huge headaches because copyright law varies from country to country; 
this has informed their redesign and their enthusiasm for RDFa.

Hope this helps explain something of where RDFa folk are coming from,

cheers,

Dan

--
http://danbri.org/

Received on Tuesday, 26 August 2008 01:58:27 UTC