W3C home > Mailing lists > Public > public-xg-socialweb@w3.org > May 2010

Re: Design rationale behind the current schema for Facebook Open Graph Protocol

From: Dan Brickley <danbri@danbri.org>
Date: Sat, 1 May 2010 23:50:38 +0200
Message-ID: <y2seb19f3361005011450r2a51c651l7e8a7f95e9b12b22@mail.gmail.com>
To: Peter Mika <pmika@yahoo-inc.com>
Cc: Fabien Gandon <Fabien.Gandon@sophia.inria.fr>, Ivan Herman <ivan@w3.org>, David Recordon <recordond@gmail.com>, "public-xg-socialweb@w3.org" <public-xg-socialweb@w3.org>, Tim Berners-Lee <timbl@w3.org>, Marie-Claire Forgue <mcf@w3.org>, Murray Maloney <murray@muzmo.com>, Richard Cyganiak <richard@cyganiak.de>, RaphaŽl Troncy <raphael.troncy@eurecom.fr>, "jacobi@csail.mit.edu" <jacobi@csail.mit.edu>, Bernard Vatant <bernard.vatant@mondeca.com>, Toby A Inkster <tai@g5n.co.uk>
On Sat, May 1, 2010 at 11:35 PM, Peter Mika <pmika@yahoo-inc.com> wrote:
> Hi Fabien,
> Thanks for encoding our agreement from the workshop! I agree with all the
> points below.
> Two additional comments:
> #1
> To strengthen commitment, †I would still encourage us to create a custom
> XSLT that maps OGP to the desired representation in RDF/XML. I'm not sure
> RIF or SPARQL Construct is powerful enough to express this mapping, and
> almost everyone has an XSLT parser at hand.

Sounds feasible, at least for the basic form of OGP.

> Would you take a crack at it? In a similar fashion, an OGP to JSON mapping
> would be also highly desirable, otherwise implementations will diverge on
> the JSON serialization.
> #2
> The four generic attributes are required, i.e. something is not an OGP
> document unless it has all four †(per spec, but maybe David can comment).

That's the impression I got. I haven't seen any discussion of og: type
defaulting, for example.

> It seems tricky to model this without a class (what would you put the
> constraint on?), so maybe time to introduce an og:Document class as a holder
> for these properties.

This kind of constraint is not the kind that comes naturally to
RDF/OWL, which is more about saying things like "people have
biological parents" than "when you see a biological parent in an XYZ
record, we need it to have a name field filled out". There are various
efforts that do go in this direction but I'm not sure we need to be
that formal here.

I might be a minority on this point (wasn't at WWW2010) but I still
would like to explore further the use of a more thing-centric RDF
representation, ie. with subject of statements not being the page. Not
every site will want this, but others (eg. RottenTomatoes, BestBuy,
Drupal, ...) are already exploring more extensive use of RDFa and may
prefer a cleaner model. It should be possible to support (and
validate, and consume, ...)  both idioms and to leave some
level-of-detail questions to the market / publishers / grassroots /



> p.s. Tools are already diverging? It looks like IMDB has OGP data, but this
> tool listed on the OGP page doesn't find it:

No, you have time backwards here I think (fortunately :), my RDFa
discussions with David happened pretty close to launch, and some
pre-RDFa designs were apparently already being implemented. I expect
the RDFa stuff will filter through over time...


Received on Saturday, 1 May 2010 21:51:14 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:22:09 UTC