[rdfweb-dev] PPD, revisited

Leigh Dodds ldodds at ingenta.com
Mon Feb 9 14:07:35 UTC 2004


The PPD thread seems (at least to me) to be wondering 
off topic, so I'm hoping we can refocus the discussion 
back on the 2 proposals on the table:

* foaf:primaryTopic

Morten pointed out that documents with multiple topics are 
quite rare and so this seems redundant. An editor can just 
check the foaf:topic of a document, if it's a foaf:Person, then 
voila!

However foaf:topic isn't constrained in any way, so it's 
legal to have multiple topics for a document. The above 
method is therefore not fool-proof in the long term.

If foaf:topic were constrained, then we'd be good to go. 
So that ought to be considered as well.

However I personally believe it's a useful feature, independent of 
the "who is this document about?" use case, to be able to determine 
the primary topic of a document. And for this reason I'd 
prefer to see a new property defined for this purpose.

* foaf:PersonalProfileDocument

I'm in favour of this because of the facility to type rdfs:seeAlso
links:

<rdfs:seeAlso>
  <foaf:PersonalProfileDocument rdf:resource="..."/>
</rdfs:seeAlso>

A scutter or FOAF explorer is then given an expectation of 
what will be in the RDF graph resulting from traversing that 
link. We already have specific expectations of what data 
may in a graph when we fish about in it for specific properties.
I don't think this is any different.

Again, like foaf:primaryTopic I think this is a Good Thing 
even independent of the of "who is this document about?" 
use case.

It would follow naturally that the foaf:primaryTopic of a 
PPD is a foaf:Person. But there's no limitation on what 
triples may be in the data, nor constraints on syntax, 
so I don't think that syntax vs. model is a concern.

Yes, someone could extend PPD and define a syntax 
constrained version, but that's not be proposed currently. 
Whether thats a good idea or not is a separate debate.

Yes, someone can have multiple PPDs. Having an extra 
document class doesn't restrict that and usual smushing 
rules can apply to merge data. Trusting the origin of the 
data, and what that trust may be predicated on, is an separate 
issue IMO.

Yes, someone can author (foaf:made) a PPD on behalf of 
someone else. That's why we have separate maker and 
(primaryTopic|topic) properties so we can distinguish these 
cases.

Use Case? The RDF web is only going to get larger, 
therefore scutters, like HTML spiders, are going to have to 
optimise their search paths through this web. Having an 
extra type of foaf:Document will facilitate that.

* summary

Both of the proposals stand on their own merits, in addition 
to serving the "who is this FOAF about?" use case. Why not 
add them?

Cheers,

L.





More information about the foaf-dev mailing list