update: partial dublin core implementation, this list, suggestion

Dan Brickley daniel.brickley at b...
Sun Jul 2 17:20:32 UTC 2000


Last from me on this for a bit...

DC Support
----------
I just hacked stopgap Dublin Core support (in addition to existing image
stuff) into the Perl implementation (ie. the client code). The robot of
course already aggregates DC stuff and anything else you throw at it in
RDF syntax (previous disclaimer about the genid/variables/anon nodes
aside...). The CGI Perl client now looks for dc:creator and
dc:contributor arcs connecting a person (agent? org?) to a resource
(typed using the draft WordNet vocabulary in my data; this type is
currently ignored though). For each authored and contrib'd document, the
Perl front end slurps out title/description/url/etc for display, and
cross-references to the screens displaying info about the other
co-authors. Individuals are currently picked out using pre-coordinated
fake URIs (ie. a bug) but should work fine with email address or other
uniquely identifying data, once the robot's rewritten.

So for eg., I've chucked in a couple of DC records, one describing
something I co-authored with Eric Miller and others. This shows up from
both our profiles as you'd hope. This can be thought of as implying
something like a foaf:collaborator relation between the linked
dc:contributors. So it provides yet another way of navigating through
the data web...

rdfweb-dev at egroups.com
----------------------
I've just fixed the egroups setting so the archives of this list are now
public, and therefore citeable etc. Default was for group-private
archives (hadn't realised this...)

suggestion...: 'ask the web' templates
-------------
Like I need another partbaked idea for this right now. Got to get the
basics done first. Anyway this was what I have in mind for later
version: 'ask the web' query/report templates. In brief: one's
rdfwebring representation should allow you to point to an RDF query /
reporting XML file that encodes a query or queries against the
rdfwebring aggregates database. Services could then (at their
discretion) read in these (from some/all users) and
expose the results of the queries in HTML. Amongst other things, it
provides a motivation for exposing rich data to the Web in RDF, and
further decentralises the design of rdfwebring facilities...

Context:
The principle behind the RDFWebRing system is "Ask the Web / Tell the
Web", ie. if some piece of data isn't private, then make it public
properly so machines can help us use it. So the system needs to be
content-neutral and decentralised, to avoid restrictions on the sorts of
things you can ask or tell the web. Right now, the robot(s) that harvest
this stuff are content-neutral, with their own traversal and storage
policies (eg. scutter gullibly slurps everything it finds). The RDF
query APIs are content-neutral, but the Web-based client(s) are
vocabulary specific. This is necessary, since general purpose interfaces
for arbitrary data tend to suck (unless you like flying through 3D node
and arc diagrams). To democratise the UI, and allow people to associate
views into the rdfweb with their online profile, we need some way of
sharing formatting templates that embed RDF queries. I'm getting
ahead of myself here: rdfq report templates are also needed anyway
since writing RDF-to-HTML code in Perl gets a little tedious (esp. if
you then have to rewrite it in Java...). I think we can count on some
such technology evolving; this is a suggestion for deploying it in a fun
way...

Scenario:
Say Eric wants to create a view of the homepages, technology interests,
publications and working group membership of all dublin core
implementors. If they're represented in the dataset, we could imagine
some RDF 'database report' template format that mixed RDF queries with
HTML, much like people use ASP/JSP/XSLT. Assuming (!) that such a
technology exists or can be prototyped adequately, here's how it should
work. Eric creates an RDF 'view file' defining what he wants to ask the
Web, and how to format the output. He links this from his rdfwebring
self-description. Subsequently, certain generous services might run his
query against their (?subset of) the rdfwebring data, and link the
resulting HTML report from their representation of Eric. Similarly,
you'd go to my 'useful views of the rdfweb' screen and see reports about
the RDF Interest Group membership, or the Bristol geek scene, or a
list of the favourite songs/movies recently recommended by my friends etc...

Definitely one to do later; lots of scope for denial of service attack,
being sued, all sorts of horrors. But this is what I see rdf being for,
so I reckon its worth building to find out for real what the issues are :-)

--danbri




More information about the foaf-dev mailing list