[rdfweb-dev] Time's running out?

Dan Brickley danbri at w3.org
Mon Aug 11 22:49:26 UTC 2003


* Victor Lindesay <victor at vicsoft.co.uk> [2003-08-11 22:21+0100]
> 
> 
> > -----Original Message-----
> > From: Jim Ley [mailto:jim at jibbering.com] 
> > Sent: 10 August 2003 01:45
> > To: Victor Lindesay; rdfweb-dev at vapours.rdfweb.org
> > Subject: Re: [rdfweb-dev] Time's running out?
> > 
> > 
> > "Victor Lindesay" <victor at vicsoft.co.uk>
> > > As the domain of foaf:knows is declared as foaf:Person, all 
> > subjects in
> > > statements with a foaf:knows (or any FOAF property with a domain of
> > > foaf:Person) predicate can be assumed to be of type foaf:Person.
> > 
> > Not assumed to be they are, or the data is a wrong.
> 
> What is the point of this remark? I have expressed myself correctly. I
> post on this list to talk about RDF and FOAF, not English. 

I found your wording perfectly understandable.

> > Why do you care about the type at all?
> 
> Because I want to write software that processes data correctly. You
> might not care about crap data but I do.

I think we all care about crap data, and are just coming at this from 
different traditions. The ommission of an rdf:type property with value 
foaf:Person isn't a perfect indicator of crapness though, nor is its 
presence a guarantee of accuracy. However tightly we define our schemas 
(RDFS, OWL etc) and document formats (DTDs, XML Schemas etc) there will
be plenty of opportunities for crappy data to slip through the gaps, 
un-noticed by machines. Take foaf:Person for example. We don't say 
whether the person has to be alive, dead, real, imagined etc. If we
spent weeks nailing down a far more ambitious version of FOAF which did 
make such distinctions (eg. INDECS, the rights description language,
had a good stab at this) we would still have a tricky time trying to 
machine-check data to see whether it met our new schema. 

Vocab-based crapness detection will only take us so far; after that,
we're into the murkier world of trust metrics, hacks and heuristics...
(which is where the pre-XML search engines live, fwiw...)

Dan



More information about the foaf-dev mailing list