Re: DC Terms, FOAF etc

Phil Archer wrote:
> In a comment on our latest public drafts [1], Ivan pointed to the 
> updated namespace for Dublin Core. Everything is now in 
> http://purl.org/dc/terms/ with legacy support for the older namespaces.
> 
> OK, in the updates I'm working on right now, I'm updating that namespace 
> in all our examples. It's particularly relevant for dcterms:issued. 
> Shout if you think we shouldn't do this.
> 
> But... looking at this reminded me that in an earlier e-mail exchange (I 
> think it was just to me, not on an archived list) he also queries our 
> use of FOAF when:
> 
> 1. DC now includes an Agent class and dcterms:creator can point to that 
> class in much the same way that foaf:maker points to a foaf:Agent
> 
> 2. FOAF, for all its popularity and usefulness, is controlled by two 
> individuals, not by an organisation. However illustrious those 
> individuals may be (and they don't come much more so than Dan and 
> Libby), this is a potential weakness in the POWDER spec.
> 
> Hmmm... looking closely at the (updated) DC vocabulary [2] it is perhaps 
> tempting to ditch FOAF and just use DC. But I'm inclined not to do so 
> for several reasons and I thought it best to record them in public for 
> possible future reference:
> 
> 1. DC Terms has an Agent class but, unless I've missed it despite 
> looking, none of: name, nick, homepage etc. To take my own organisation 
> as an example, I use those three and more in our FOAF file [3] and I 
> believe that, whilst FOAF is not a perfect fit, it's very close to it 
> for our uses cases.
> 
> 2. DC is massively successful in the publishing world (its origin) and 
> FOAF is massively successful for linking people and organisations. I 
> know Dan is trying to find a way to put FOAF on a more stable footing 
> but from an outside perspective, it's an established vocabulary and the 
> one people expect to see used. RSS isn't in the purview of an 
> organisation either - but it's a method of doing things that crops up 
> occasionally ;-)
> 
> We have talked about this within the group. I recall Kai raising the 
> issue ages ago, for instance, but the feeling has been that we'd like to 
> stick with FOAF. If the Rec Track process or the weight of comments 
> received makes it clear that we should not mandate its use, OK, we have 
> a get out, but personally, I'd like to keep things as they are: dcterms 
> for issued and FOAF for the the attribution info.
> 
> That said, if the WG wants to use dcterms:creator->dcterms:Agent rather 
> than foaf:maker -> foaf:Agent - it's not /that/ much work to do it.

Thanks for the nice note. In the interests of simplicity and POWDER 
users not needing to include loads of extra namespaces to express a 
basic label, I do suggest making use of the dcterms vocab instead. The 
new Dublin Core stuff is much closer to a FOAFish design and we'll 
probably have direct mapping/equivalence expressed in the schemas at 
some point (so long as that doesn't mess up our OWL 2 compatibility plans).

I do think we have some stuff in FOAF that adds value to POWDER labels, 
and we'll be having more yet. For instance, isn't it rather nice to know 
the :opend and :homepage of a label creator? Or identify them indirectly 
via a hashed mailbox ID that can be fed to Google's SGAPI service, 
http://code.google.com/apis/socialgraph/docs/ ?

So my recommendation would be to make sure the basic creator/Agent thing 
is doable in plain DC terms, but allow FOAF for adding more optional 
detail.

cheers,

Dan

ps. cc:'ing Edd, with whom I was recently discussing this

> Cheers
> 
> Phil.
> 
> 
> [1] http://lists.w3.org/Archives/Public/public-powderwg/2008Jul/0026.html
> [2] http://dublincore.org/documents/dcmi-terms/
> [3] http://www.fosi.org/fosi.rdf
>

Received on Tuesday, 8 July 2008 12:40:31 UTC