Ivan’s private site

December 7, 2008

RDFa as an RDF serialization syntax

Just a small thing, really… On a blog Sebastian Heath announced that the Greek, Roman and Byzantine Pottery at Ilion (GRPBIlion) database exports its content as Linked Data using RDFa: see the HTML page itself or its RDF representation via RDFa Distiller. Mark Birbeck already blogged about this on RDFa’s blog site.

What caught my attention, however, is a question on the original blog by Sergey Chemyshev who asked:

Is there any reason why you used XHTML+RDFa instead of RDF/XML or N3 for this? It looks like http://classics.uc.edu/troy/grbpottery/database.html doesn’t have any HTML for presentation purpose, just to produce RDF triples – what was your reasoning for embedding them into HTML?

Of course, this is a valid question but… isn’t the answer “so what” (no offense to Sergey!)? What I mean is: for most of the RDF data around RDFa is a valid serialization just as RDF/XML or Turtle are. Of course, the main message around RDFa is that it gives you tools to add RDF data to an (X)HTML page. But we should realize that one can look at it in a different way, too: it is a tool to serialize your RDF graph so that it can be displayed in a human readable form easily via a browser. The HTML side of is not necessarily fancy; it can be, actually, very simple as indeed Sebastian’s page is. But it is readable for humans, certainly more readable than RDF/XML or even Turtle would be. And I think that may be enough of a reason to use RDFa in such circumstances, too!



  1. So what? What about consumers that don’t grok RDFa? That page doesn’t have a profile which is GRDDL friendly, so that is even more lost possible means to get out RDF.


    Comment by Daniel O'Connor — December 8, 2008 @ 3:03

  2. On long term, various RDF consumers (databases, SPARQL endpoints, etc) would consume those files just as easily as they do GRDDL profile-d files. (I agree that adding the RDFa profiles to pages, though not required, is helpful in this respect).

    By the way, setting up Apache to get to the RDF content of such files automatically is also possible, see my earlier blog: https://ivanherman.wordpress.com/2008/04/27/setting-up-and-rdfa-file-with-apache-second/. Yes, it requires some access to apache, but then the choice between the HTML or the RDF content is a matter of content negotiation that all RDF consumers, afaik, do already…

    Comment by Ivan Herman — December 8, 2008 @ 15:24

  3. In response to your “so what?” Ivan, I would argue that there is a separation of concerns problem with RDFa used this way. It mixes human-readable and machine-readable representations into one overly generic bundle. For example, what would be the MIME type of such a serialisation format? I guess application/xhtml+xml would be the most likely candidate, but that signifies nothing to clients about the richer semantics actually embedded in the doc. Surely separate human-readable (i.e. application/xhtml+xml non-overloaded XHTML) and machine-readable (i.e. application/rdf+xml) representations of the same resource would be a much cleaner and also more web-friendly approach?


    Comment by Julian Everett — December 9, 2008 @ 15:49

RSS feed for comments on this post.

Blog at WordPress.com.

%d bloggers like this: