RDF hacking ILRT Home

RDF: Forward Knowledge Exchange

Author: Dan Brickley
Date: Nov 2000
Latest version: http://ilrt.org/discovery/2000/11/fk/

Status of this Document

Rough draft, in progress.


This document describes a system that uses XML/RDF to characterise the contents of networked databases.


This paper describes a simple RDF vocabulary for providing query-routing facilities between Internet-based searchable information systems. The proposal presents a mechanism which can use the XML serialisation syntax for RDF to create "bulk metadata" records that describe the contents of searchable networked databases. This work is heavily influenced by the centroid mechanism defined in the WHOIS++ directory protocol [WHOISPP], and by subsequent activities within IETF's CIP group which have abstracted the 'forward knowledge' idea away from an association with particular search protocol [see KIRRIEMUIR]. The FK vocabulary should be considered complementary to these efforts. Whilst this document focusses on an XML representation of the RDF models it defines, those same models could equally be mapped into CIP-based representations.

The FK vocabulary adopts the Warwick Framework [WARWICK] approach of modularising complex resource discovery problems and tackling the parts separately. Consequently, this document only presents a vocabulary for exchanging "bulk metadata" records describing the summarised textual contents of a searchable internet resource. The problem of describing the semantic coverage of a resource using other metadata vocabularies, intellectual property rights, access permissions, machine interfaces and so forth is a problem which can be tackled in another namespace. The scope of the FK vocabulary, broadly, is as follows: describe the public names for the fields in some searchable resource, then list all the unique words that appear in each of those fields. Vocabularies such as the Dublin Core [DC], the Platform for Privacy and Preferences [P3P] and others will supply part of the solution to other aspects of the problem outlined here.


Finding resources on the internet is easy; finding the right resource is not so easy. There is always more than one place to look, and always more than one way of looking. Information-retrieval and directory protocols such as Z39.50, LDAP and WHOIS++ have made it feasible to search multiple collections of resources

By using RDF, we can draw a clear distinction between the logical structure of the stored data and the particular serialisation syntax used to ship those RDF statements around the network. This document uses the popular XML format for the concrete representation, adopting the RDF syntax specified in [RDFMS]. It should be noted that non-XML mechanisms for sharing RDF forward knowledge may become available. In particular, the ability to access remote RDF databases through some form of distributed objects system could bypass the need


NOT just (or even) for RDF Query. Different query languages / models could have their own FK 'information set', and could perhaps share these, or have own interpretation of a standard model (eg. DC properties).


A writeup of ideas from ROADS, DESIRE, IMeshTk...



Maintained by: Dan Brickley