- From: Niklas Lindström <lindstream@gmail.com>
- Date: Sun, 27 Nov 2011 13:11:30 +0100
- To: Toby Inkster <tai@g5n.co.uk>
- Cc: Ivan Herman <ivan@w3.org>, Shane McCarron <shane@aptest.com>, public-rdfa-wg@w3.org
Hi Toby, 2011/11/13 Toby Inkster <tai@g5n.co.uk>: > On Sun, 13 Nov 2011 14:18:21 +0100 > Niklas Lindström <lindstream@gmail.com> wrote: > >> > A good RDF predicate is one that fits naturally into the following >> > two phrases: >> > >> > "has ..." >> > "is ... of" >> >> Actually there are a number of common predicates which don't follow >> this pattern (of a property noun) but rather a verb. For instance in >> Dublin Core there is 'replaces', 'requires', 'references' and >> 'conformsTo'; in FOAF there is 'knows', 'depicts' and 'based_near'. > > FOAF has very badly named properties. It has no consistency. Witness > camelCase versus underscore_seperators. It is too ingrained to change > though. Most of the original Dublin Core properties were quite good; > some of the newer ones less so. > > The role/noun pattern (or has/of antipattern) if you prefer is an > established meme - not just some idea I've dreamt up. > > http://www.w3.org/wiki/HasPropertyOf > http://www.w3.org/wiki/RoleNoun Absolutely. I know you're well versed in vocabulary design and take your points very seriously. And I pretty much agree. My main point though was that I think "uses" was better than "has" *if* any one of them was to be in the name of the property. (And in general I believe there are cases where "<verb>" works better than "has <noun>" for property names. But that's another debate.) So back to the issue: I'd also very much prefer just 'vocab' (<http://www.w3.org/ns/rdfa#vocab>). I'd welcome a quick vote on changing that. >> Do you have an opinion regarding whether the object should be an IRI >> or a literal? > > Purely from a semantic point of view, an xsd:anyURI typed literal > probably best reflects reality. But from a pragmatic point of view, an > IRI is probably most useful, as it allows "joins" between the document > being parsed and the vocab's graph in SPARQL. I'm not especially > bothered either way. Good points. I can see why that could be useful. And I suppose given that the vocab value in practice is only used for *dereferencing* something, and not any string processing, nor to be followed logically, and IRI should work (even if it may refer to an ontology information resource only indirectly, relying on HTTP mechanics). Just as long as everybody is clear on that, I'll rest my case. Best regards, Niklas
Received on Sunday, 27 November 2011 12:12:31 UTC