Documenting changes

Since the last posting to this list there have been some changes to the 
RDF-Content Labels spec so I should outline and justify those changes in 
public, so to speak.

There are two basic Classes: Content Label and Ruleset

A Content Label is a container for descriptors. A resource can point to a 
Content Label as being a description of itself. This is true of RDF in 
general of course but the underlying point of RDF-CL is that the 
descriptions are drawn from a particular area of interest, such as child 
protection, and are not designed to provide a full description of the 
resource. Whether the author is Tom, Dick or Harry, and whether the subject 
matter is nuclear physics or Scrabble scores, the same RDF-CL can still be 
applicable because it says that the background colour is white.

The Ruleset Class can be used to identify which label applies to which 
resource based on its URL which gets put though a sequence of rules based on 
Regular Expression matches.

The Ruleset now sets up a series of "Global variables" about the domains 
covered. These are the hosts themselves and any path restrictions plus the 
default label. The rules that follow may then override that default label.

It became clear that as well as Content Labels that gave detailed 
descriptions, it was also desirable to have two other types of label. Fist 
of all, one for management information (somewhere to put your Dublin Core 
and Creative Commons stuff). Management Info can be set independently of 
other labels so that, to extend the silly analogy above, those 500 white 
pages and 3 blue ones are all copyright John Smith Colour Ltd.

The other type of label is a classification. It may be while,l it may be 
blue but the subject mater is still Japanese Politics, or the genre is "18th 
century romance" etc. (film classifications like PG-13 go here too). 
Classifications are defined as being a Class and would not normally be 
parsed for properties in the way that a Content Label would be.

So we have some properties emerging from this to link Rulesets to Content 
Labels

hasDefaultLabel
hasDefaultClassification
hasDefaultManagementInfo
hasLabel
hasClassification
hasManagementInfo

(the latter 3 appear in rules that override the first three that are set as 
"globals" in the Ruleset.

NB. We need to look at implementations in RSS. At present those properties 
are defined as having a domain of label:Ruleset but we should be able to set 
a default label at the channel level in an RSS feed and then override it 
with  label:hasLabel if necessary on he individual item. Then there's RSS 
2.0, Atom and all that to get into. A 2 or 3 pipe problem Dr Watson.

Finally in this brief briefing, the way hostRestrictions are encoded changed 
a bit too. We wanted it to be possible to create a single RDF instance with 
all the label data in it but to keep the hostRestrictions external. This is 
so that a content provider who routinely creates sites on new domains can 
essentially put an API on their database of domain names to return an RDF 
fragment that lists those domains. Hence there's a specialised class called 
Hosts.

The next issue I want to address is how to include one label within another. 
Having gone to all the bother of going against the RDF philosophy and 
setting up the system so that one label can override another, we need a way 
to explicitly merge them! A new property of label:include with a domain and 
range of Content Label seems a likely way forward.

ICRA has now switched over to RDF-CL from PICS so I can point you to some 
examples of their usage:

www.aol.co.uk includes a link to http://www.aol.co.uk/icra/labels.rdf
www.gsmworld.com links to http://www.gsmworld.com/labels.rdf
And here's a site that already had an RDF file to which they just added an 
ICRA label:
www.six27.com links to http://www.six27.com/sitewide.rdf

RDF-CL is part of the Quatro project http://www.quatro-project.org/ that has 
defined a vocabulary for trustmarks 
http://www.quatro-project.org/vocabulary/1.0/ for which the namespace is 
www.purl.org/quatro/elements/1.0/

It is a candidate technology for the Mobile Web Initiative's mobileOK 
trustmark. See http://www.w3.org/2005/MWI/

RDF-CL use cases: http://www.w3.org/2004/12/q/doc/rdf-contentlabels.html
RDF-CL schema description 
http://www.w3.org/2004/12/q/doc/content-labels-schema.htm

Phil Archer

Received on Monday, 8 August 2005 14:55:54 UTC