Re: Google adds JSON-LD support to Gmail

On May 16, 2013, at 4:47 PM, Dan Brickley <danbri@danbri.org> wrote:

> 
> On 16 May 2013 13:51, Dan Brickley <danbri@danbri.org> wrote:
> >
> > On 16 May 2013 07:08, "Manu Sporny" <msporny@digitalbazaar.com> wrote:
> >>
> >> Google just added JSON-LD support to Gmail. It was announced at Google
> >> I/O (thanks to Bob Du Charme for spotting it):
> >>
> >> https://developers.google.com/gmail/schemas/embedding-schemas-in-emails
> >>
> >> This is a pretty big validation of the technology. Great work to
> >> everyone that has been involved in the grueling process of getting the
> >> specs to where they are today! We should all be very proud of this
> >> breakthrough!
> >>
> >> They have decided to not use a URL in the '@context'. I'm currently
> >> trying to find out why they did this. If they did this on purpose, we
> >> may have to do another LC to accept 'keywords' in the @context field.
> >
> > Happy to discuss this in detail after Google IO finishes i.e. next week.
> >
> > Briefly, this was all about simplicity for partners...
> >
> >> Ideally, we can just ask them to put 'http://' in front of the
> >> parameter. It's the only place that they deviated from the spec (which
> >> is actually pretty good).
> 
> A couple of points on this:
> 
> 1. We (Google) can parse this if written @context="http://schema.org" and we'll find a way to document that.
> 
> 2. We'd also like to start a conversation about allowing the simpler, shorter form by defaulting to http:// if not present.
> 
The value of @context may be a relative IRI, so seeing "schema.org" would otherwise be interpreted as being relative to the document location. This is a useful pattern, as context documents are often co-located with the base document, or even against other context documents. For example:

base document:
    {
      "@context": "/contexts/mycontext.jsonld",
      "@id": ""
      "@type": "Person"
    }

at /contexts/mycontext.jsonld on the same server:

    {
      "@context": {
        "@vocab": "http://schema.org/"
      }
    }

Are you suggesting that we make a special case for "schema.org", or abandon the use of relative IRIs for remote context documents? Of course, as we just had our second last call, this would require that we go to a third. Is adding "http://" in front of "schema.org" really that bad? It is required in RDFa and microdata.

Gregg
> Dan
>