Re: what's the subject?

Hi,

Markus and Tomasz, thank you!

Le 16 mars 2015 à 08:52, Markus Lanthaler <markus.lanthaler@gmx.net> a écrit :

> On 14 Mrz 2015 at 19:41, Tomasz Pluskiewicz wrote:
>> On 2015-03-13 18:44, Markus Lanthaler wrote:
>>> 
>>> Shouldn't the graph be about the resource that was used to retrieve it?
> In
>>> other words, doesn't the URL tell you what the main resource is? I do
>>> understand that hash URLs etc. complicate this.
>>> 
>> 
>> Hash or not I think it is a real issue. Another example is when a
>> representation is split across multiple urls. Think </bob>, which returns
>> 
>> </bob> a foaf:Person ; foaf:givenName "Bob" .
>> 
>> and </bob/friends> which returns
>> 
>> </bob> foaf:knows </alice>, </frank>, </mike> .
>> 
>> This is especially true for anyone trying to enhance existing JSON API
>> so that it becomes JSON-LD: The requested URL </bob/friends> does not
>> have to be an actual resource identifier but merely an address for
>> additional data about </bob>. This is how many APIs are structured and
>> it is indeed an obstacle with RDF Linked Data where we struggle to make
>> each URL an identifier.
> 
> I see.

cool!

(Markus, I'm late in answering your previous message on this topic, sorry). 

Other use case: when content can be accessed through several URLs, but there is one "canonic URI".
"semantic web people" use case: linking the "page" and the "real world thing"

A statement using this property, inside the returned graph, is metadata about the graph. It is useful to have it included inside the returned data : things are explicit, data is self contained. That's easier for the client, be it a "headless" one or not.

Example of simple application that it allows: indexing set of graphs by "main subject". Aggregation of data based on common (or related) "main subject"

> I'm not sure though Hydra is the right place to define such a
> property. Are you aware of the related discussions in Schema.org land?
> schema:mainEntity, see
> 
>  https://github.com/schemaorg/schemaorg/issues/301
> 
> What we could do, if this proposal gets accepted and people think it would
> be useful, is to include an example explaining it to our spec.

that's fine for me. The point is that such a property is necessary (at least, I needed it in most if not all of my LD stuff). It's good if data publishers can rely on one recommended property.

> 
> 
> Cheers,
> Markus
> 
> 
> --
> Markus Lanthaler
> @markuslanthaler
> 
> 
> 
> 

Best,

fps

Received on Monday, 16 March 2015 09:04:01 UTC