- From: Benjamin Nowack <bn@kasabi.com>
- Date: Sat, 22 Oct 2011 16:35:12 +0200
- To: Jeni Tennison <jeni@jenitennison.com>
- CC: public-html-data-tf@w3.org
On 10/21/11 8:57 PM, Jeni Tennison wrote: > On 21 Oct 2011, at 18:03, Benjamin Nowack wrote: >> Looking at the influence angles this TF has on RDFa vs. MD editors, >> I wonder if it might be worth looking in the other direction, too, >> to check if some of RDFa's features could be directly applied to MD >> if certain criteria are met. > [...] > First, wouldn't it mean that microdata-unaware RDFa parsers would > generate some very different triples from the pages than the > microdata-to-RDF parsers? @vocab, @datatype, and @typeof as MD-additions shouldn't have practical side-effects to RDFa parsers that don't support MD (well, a couple of typed bnodes, which should be semantically ok). So far, there aren't any "official" MD2RDF parsers, so implementers could still adapt their code. But you are right that this would need some work to verify that real-world examples really can't trigger unwanted triples. >The only way I can see to avoid that would > be to add @about attributes into the page. The moment we add @about, we may trigger the creation of triples by RDFa processors. I was more thinking about using these RDFa features in contexts where a non-MD-aware RDFa parser doesn't create triples anyway. > If you were dealing with > coexisting microdata and RDFa anyway, adding extra types about an > entity through RDFa would be easy enough without changing anything > about how the microdata itself were processed. I wouldn't think of this as using RDFa really, it would be something like MD+, where the "+" just happens to be syntactically identical to RDFa. The additions would have no effects on pure MD parsers, and no effects on pure RDFa parsers, but could address the needs of RDF-targeting authors and related MD/RDF parsers. > Second, the vast majority of publishers aren't going to give two > hoots about how RDF processors see their data, so they're not going > to add things like vocab or datatype. Yeah, but they wouldn't need to be aware of RDF, e.g. Martin could just say that GR authors should but the GR types into @typeof and can keep the schema.org ones in @itemtype. >I think we need a mapping that > will work even when they don't, and if we have *that*, there's no > need for *anyone* to add those attributes. That would of course be desirable. I think the central issue is generating predicate URIs from short names in MD in a generic way without violating either the MD model or RDF principles (where you shouldn't make up URIs in other people's URI space). > Does that make sense? Yes, it does. :) > Are you starting to work on parsing microdata into RDF now? Yes, slowly. Just came across that unexpected situation where MD is more granular than RDF, as you can (or even have to) keep the vocabulary separate from the property name. It's sort of a dead end for the RDF conversion process if properties aren't URIs. Cheers, Benji > > Cheers, > > Jeni
Received on Saturday, 22 October 2011 14:35:42 UTC