W3C home > Mailing lists > Public > www-rdf-interest@w3.org > November 2002

Re: Announce: Browse RDF like the web

From: Damian Steer <dsteer@hp.com>
Date: Sat, 09 Nov 2002 01:53:17 +0000
To: Paul Prescod <paul@prescod.net>
Cc: www-rdf-interest@w3.org
Message-ID: <m2el9vcxv6.fsf@rudiger.danbri.org>

Hash: SHA1

Paul Prescod <paul@prescod.net> writes:

> Damian Steer wrote:
>> Well I'm flattered that you're offering to host this service. Thanks
>> very much. :-)
> I don't know about the goals of your particular project but I could
> see some interesting marketing possibilities if HP owned the site that
> became the "user interface" to the Semantic Web. Perhaps it would
> evolve into Yahoo or Google of the SW. I'm just throwing out ideas,
> you obviously have a better idea of what you are trying to achieve!

Forwarding you thoughts to <carly@hp.com> :-)

Yes, I think I should explain what this project is about:

Well, there is rdf data out there and available, from a variety of
sources (RDF/XML documents, TAP, Sesame, Squish SOAP services, etc etc
etc) using a variety of schemas (DC, FOAF, vCard, hybrid calendar, etc
etc). And (hopefully) they'll be more.

The 'goal' was to just make a browser. Something that could use all
those sources and make the data presentable. Why? Well there is useful
data available. I also find it quite hard to read RDF/XML, N-Triples
and N3 for even moderately complex data sets, and that makes tracking
down bad data hard. I'm sure people will find other uses.

You can alternatively view the project as an experiment. RDF data is
semi-structured, and can sprawl over multiple sources. The model part
of brownsauce tries to cope with this - the HTML renderer is just one
possible (imperfect) view. Dealing with the semi-structured aspect
first, I tried a number of methods to delimit meaningful sub-graphs,
with only knowledge of RDF and RDFS. Sometimes it's obvious: just a
node and it's neighbours. But then you find things - like vCards,
events, RSS channels - which are a great deal more difficult. Adding
in knowledge of these known patterns doesn't necessarily help since
nothing prevents <randomNamespace:property> cropping up. On the
aggregation side you have discovery mechanisms like seeAlsos. (Indeed
I hope the browser will encourage the use of seeAlso).

Bit of a brain dump, but anyway that's roughly how the project's
worked so far.

> Another freely offered idea: if particular types of data can look bad
> under the particular "semantic stylesheet" you're using, then maybe it
> should adjust its "semantic stylesheet" by RDF schema, as is common in
> the XML world (where an XSLT is associated with a particular
> vocabulary).

The 'semantic' part of the styling is no more than minor tweaking for
the delimiting algorithm - eg custom.rdf contains statements like
rdf:Seq rdfs:subclassOf bs:Traversable (ie keep going over Seq nodes)
- - plus some useful things to augment schemas - eg I made foaf:name a
subproperty of rdfs:Label, again in custom.rdf. The 'bad' data usually
stuff I haven't tweaked enough.

Much of the 'semantic styling' is simply proper use of rdfs:label and
sensible modelling (like not using URIs for things which don't
need to be globally identified).

> Also, do you think it would be practical to have an equivalent tool
> that ran as an applet? I'm tempted to fiddle with it and see if it can
> be applet-ized but if you know in advance that it isn't practical then
> I wouldn't bother.

Hmm. Well, the model is nicely separate from the display, and pretty
simple; however my concern would be the size of Jena and Xerces, and
the need to load all the schemas referenced by the data. But people
use mozilla, and that's not exactly light weight :-) Of course Mozilla
has RDF support built in, so maybe a tool could be built within it?
Out of my depth there.

An alternative option is one Jim might like for foafnaut. Make
brownsauce a (RESTful) service which returns easy-to-parse data. The
applet could use that service to do some of the work.

Went on a bit, and it's getting late. Hope that all made sense and
clarifies the project a little.


Version: GnuPG v1.0.7 (Darwin)
Comment: Processed by Mailcrypt 3.5.6 and Gnu Privacy Guard <http://www.gnupg.org/>

Received on Friday, 8 November 2002 20:53:20 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:07:43 UTC