[rdfweb-dev] PPD

[Morten Frederiksen]

On Wednesday 04 February 2004 12:29, Dan Brickley wrote:
> 1) note that any RDF/XML document can draw upon FOAF vocab to
> describe people, including but not limited to its author, publisher etc.
No argument here.

> 2) in the user world, there is a colloquial notion of 'my FOAF file'
> which has no machine-detectable representation. This has proved
> problematic for sites like PeopleAggregator, as they wanted to be
> able to figure out, given a "FOAF file URI" whose the "FOAF file
> of".
I can see the point, but I don't see where foaf:topic falls short.
If the file is "my FOAF file", in the above definition, then I can't see how 
it can have more than one foaf:topic, me.

> TypePad and others similarly have heuristics which exploit
> expectations in this vein. PPD would make such assumptions more
> explicit: some (but not all) RDF/XML files that use FOAF are
> 'Personal Profile Documents', ie. made by, and primarily about, the
> person they're profiling.
I think it's a mistake to confuse who made a file and who it's about, that's 
two different issues (hence the need for two different properties in the 
first place).

> 3) foaf:made's "I made this" isn't enough, since you can make RDF/XML files
> which use FOAF vocabulary, yet have them mostly be non-FOAF, and have them
> describe other people (eg. I publish the FOAFCorp dataset).
Agreed.

> 4) "I'm the foaf:topic of this RDF/XML" doesn't quite cut it either; there
> are many ways an RDF/XML document could have me as (one of possibly
> many) of its topics. I might be a fatcat director featured in
> foafcorp.rdf, for example.
That's quite allright (even if I have yet to see a document with multiple 
foaf:topic's), but then it is no longer a PPD anyway?

By the "definition" of a PPD, it's a document "about a single person", so if 
a document has more than one topic, how can it still be a PPD?

> 5) we have currently no way of asking "how many FOAF files are there".
> We can ask how many RDF files use FOAF vocab, and that's about it.
> It surely would be nice to have a growth chart showing the number of
> documents of this type.
Hmm, yeah, I guess, but this can be done right now simply by counting 
documents that have a foaf:topic of a person...

> PPD would be a subclass of foaf:Document. We could see other
> subclasses of foaf:Document defined, in FOAF or elsewhere, that capture
> specific FOAF-based file formats or graph structures, ie. constrained
> either in terms of concrete syntax or in terms of the RDF data carried
> by that syntax. For eg the profiles Leigh sketched (using Schematron,
> initially) at http://www.foaf-project.org/2003/schemas/profiles/readme.html
> would correspond to different classes of RDF/XML document. Unlike these,
> PPD doesn't currently set any machine-checkable (ie. hard and fast)
> rules about what it'd contain. But the general expectation that it would
> be foaf-a-matic-esque seems reasonable.
I think this is where the dog is buried, so to speak, as I don't think 
profiling documents this way is the way to go.

Call me naïve, but if FOAF as RDF is going to survive, we have to keep 
insisting on the model, not the syntax. By restricting the syntax, we may 
help out a lazy app-developer here and there - on the short term, but it will 
stop the progress towards more RDF - on the long term. We'd essentially be 
doing them (and us) a disfavour.

> > >property: foaf:mboxsha1sum (clarification)
> My inclination is not to normalise
> forcibly, but insert a 'best practice' note into the spec encouraging
> hostnames to be written in lowercase if there is no over-riding reason
> to do otherwise. But I've no strong opinion on this.
That seems like a good idea, it should of course be mentioned with foaf:mbox 
as well.


Regards,
Morten