The (woefully incomplete) general
overview provides more detailed explanation of what we're
trying to do.
RDFWeb is an experimental linked information system, where the linked items are machine-understandable documents in the World Wide Web. These documents, written in the XML, RDF and XHTML languages, are harvested by our Web indexing programs, which then merge the resulting information to form one big database. By exploiting various features of the RDF technology alongside related tools such as digital signature, we can use the Web to retrieve structured information that machines can process and act upon. RDFWeb is best understood as an exploration of the question: "what would it be like if machines could read what we say in our Web homepages".
See the sitemap for links to experimental implementations, research notes etc.
The initial RDF/XML vocabulary that underpins RDFWeb is called FOAF ("Friend of a Friend"). It is used alongside other vocabularies (RDFCore, Dublin Core, Wordnet, RSS 1.0) to provide mixed-vocabulary descriptions of people, organisations, web services and other information resources.
RDFWeb is constructed from a number of services. The basic building blocks are RDF and XML parsers, and XSLT transformations: this technology is available "off the shelf". RDFWeb provides authors of such software with a large, distributed and heterogeneous collection of test files to help them refine their software. In addition to parsers, RDFWeb agggregation points require some sort of RDF database support. Again, these aren't a core part of RDFWeb itself, the idea is to use RDFWeb to help us develop and test such tools. The most visible component of RDFWeb is the Web-based user interface that presents the aggregated data back to users. We have had a Perl and a Java implementation, currently only the latter is maintained.
RDFWeb raises some interesting issues relating to trust and provenance. We have created a simple RDF vocabulary, WOT ("web-o-trust") to test some ideas about the combination of digital signatures with RDF content. Currently, some RDF files in the RDFWeb have been signed using GPG or PGP; those files typically contain a pointer to the result of that signing, and may also contain or point to information about the signer, and claims about reciprocal key-signing events. The intent here is to provide enough meta-information to support a variety of trust and provenance strategies.