RDFWeb Scutter HOWTO / developer roadmap

Author: Dan Brickley
This Version: $Id: xml-12.html,v 1.9 2003-02-12 13:23:35 danbri Exp $ (cvs history)
Nearby: ScutterSpec, FOAF, foafnaut.org

(tambien disponible en español; traducción por Leandro Mariano López / inkel)

12 Step Programme

1. Load some RDF file from URL into an RDF database
2. ...and check for rdfs:seeAlso properties
3. ...and load those files too (and repeat, eg. twice)
4. Use some RDF API to make HTML/Web interface to this data, eg. representing people, documents, organisations, photos etc. described in the RDF.
5. Implement some 'smushing' techniques to defragment the data, ie. folding together different descriptions of the same thing by knowing which RDF properties are uniquely identifying (ie. 'UnambiguousProperties'). Encourage your RDF toolkit/database vendor or opensource project to ship such facilities as standard (showing a running app may help).
6. Implement some provenance-tracking facilities, so that for each scuttered / indexed RDF document, your database system knows how/where the data was acquired. Record the means of acquisition (URI dereference, mail mid: etc), parameters (HTTP content negotiation etc), date and time. Check digital signature of document if relevant, recording fingerprint of signing key. Encourage your RDF toolkit/database vendor or opensource project to ship such facilities as standard (showing a running app may help).
7. Improve the HTML/WWW interface to exploit any additional information this provides. Take care to offer mechanisms that allow information to be removed/hidden if people ask, and to respect relevant data protection laws.
8. Add some more gadgets to the user interface, such as foaf:humanGenomeChecksum, foaf:geekCode, foaf:archNemesis and other follies.
9. Improve the smushing to use smarter heuristics, like collections of properties which when taken together are uniquely identifying. Have the system read its config from these from the Web. Encourage your toolkit/database friends to provide these as standard, if you can show them to be useful.
10. Implement some interesting bits of the DAML+OIL or WebOnt: inverseProperty, UniqueProperty. And rdfs:subPropertyOf, rdfs:subClassOf, rdf:member etc from RDF/RDFS. Enjoy.
11. Implement substring based searching, phrase search etc., by writing out text files that Z39.50 or other digital library indexing tools can use. Integrate these as backends behind an RDF API.
12. See if it scales.