Hi Asger
 
Thanks for the clarification on these.  I thought there would be issues with Fedora's notion of "type" - so just wanted to clarify.  And I hadn't spotted that owl:hasValue wasn't in OWL Lite.  Thanks for the suggestion on the XSD alternative, that might just work!

Regards
Steve
-----Original Message-----
From: Asger Askov Blekinge [mailto:abr@statsbiblioteket.dk]
Sent: 13 December 2011 14:21
To: fedora-commons-users@lists.sourceforge.net
Subject: Re: [fcrepo-user] ECM - validating the "type" of an object

On 12/13/2011 01:51 PM, Stephen Bayliss wrote:
In ECM, the type of an object is defined effectively from its CModel.  One can validate the type of the target of a relationship, the typing being specified by the hasModel relationship.
 
Is it possible to do any additional type validation; eg if I want to enforce that all objects "belonging" to the CModel must assert an rdf:type relationship?
Sure, specify the minCardinality for the rdf:type relation. That way, all subscribing objects must have an rdf:type relationship.

 
I've tried adding an owl:Restriction on the rdf:type relationship using owl:hasValue to specify the value of the target (rather than its class) - but this doesn't look as if it is implemented.
http://www.w3.org/TR/owl-ref/#hasValue-def
NOTE: the value constraint owl:hasValue is not included in OWL Lite.


 
I've also tried using an owl:Restriction on a hasMember relationship; instead of specifying the target's CModel class I used a value that the target is itself asserting using rdf:type, but it doesn't lok like this is tested for.
rdf:type is a sore point for OWL. Fedora CMA does not concern itself with rdf:type. A fedora object is NOT "of a type" just because it subscribe to a content model.
Now, one could define
fedora:hasModel isSubtypeOf rdf:type
The problem here is that, to owl, fedora:hasModel is an owl:ObjectProperty and rdf:type is a rdf:Property. To make the above statement understandable to a reasoner, you would need owl full.

For the purpose of ECM, i defined that
object fedora:hasModel contentmodel creates a phony rdf property

object rdf:type contentmodel#class

This property is, of course, not added to the triple store.

Remember that it does require owl full, if a content model should be both an object and a class. That is why we use contentmodel#class for the rdf:type information.
 
So a question, is ECM relationships validation restricted to:
 
1) checking the cardinality of a relationship
2) checking the type of the target of a relationship where the type is specified using hasModel
 
Is there anything one can do for enforcing rdf:type assertions?
writing a schema for the RELS-EXT datastream?
<xsd:schema
        attributeFormDefault="qualified"
        elementFormDefault="qualified"
        targetNamespace="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
        xmlns="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
        xmlns:xsd="http://www.w3.org/2001/XMLSchema">
    <xsd:element name="RDF">
        <xsd:complexType>
            <xsd:sequence>
                <xsd:element minOccurs="1" maxOccurs="1" name="Description">
                    <xsd:complexType mixed="true">
                        <xsd:sequence minOccurs="0" maxOccurs="unbounded">
                            <xsd:any processContents="lax"/>
				<!-- Insert required element here...-->
                        </xsd:sequence>
                        <xsd:attribute name="about" type="xsd:anyURI"/>
                    </xsd:complexType>
                </xsd:element>
            </xsd:sequence>
        </xsd:complexType>
    </xsd:element>
</xsd:schema>

But on a more serious note, no, there isn't. To do this right, we need

1. To define the implied relation between fedora:hasModel and rdf:type
2a. Find an apache-licensed reasoner, that support more than owl-lite (or switch fedora to gpl3, and use pellet)
2b/ Or: choose to support some subset of owl full/DL specification, and thus in effect, make our own dialect of owl.

Regards

 
(Though I realise in practice, with reasoning support, one would make a statement about the CModel#class that it is eg a subtype, or owl:sameAs etc).
 
Thanks
Seve