- From: Markus Lanthaler <markus.lanthaler@gmx.net>
- Date: Thu, 3 Oct 2013 22:40:00 +0200
- To: <public-hydra@w3.org>
On Thursday, October 03, 2013 2:13 PM, Thomas Hoppe wrote: > Hi, > > I make an aggregated response here: > If the defined operations have, as you say, hardly a binding character > and > every API provider is encouraged to define his own, specific operations > then I would agree to not having any operation on the spec or clearly > marking them as examples. If everyone defines their own specific operations, that wouldn't help much. What I envision is that "standardized" operations are defined for common application domains.. think ecommerce, microbloggging, etc. > I still recommend however to resolve the CRUD wording issue. Again, if you could propose some concrete wording, that would help immensely. > Regarding the use case for a ReadResourceOperation, there is an obvious > one: > If you want to specify the returned data structures with the > corresponding attributes for read operations in the API documentation, > you need this operation to specify this. Can't you use rdfs:range for that? It tells you of what type the resource referenced by a link is. So if myproperty rdfs:range MyClass . then, when I see a statement like /entity1 myproperty /entity2 I can infer that /entity2 is an instance of MyClass. > The sentence could read: > "Hydra is not limited to public APIs but can also be used on concert > with an arbitrary > authentication solution. In hydra, the operations offered by a resource > can be embedded in the representation > which allows to specify authentication constraints implicitly on a per > request basis." What about changing this slightly to (given that it will be put in the core spec): Even aspects such as authorization and authentication are beyond the scope of this specification, it is worth noting that the operations and links offered by a resource can vary per client (sorry, I don't find a better word at the moment). An anonymous client might only get a subset of the links and operations served to a fully authenticated client. What do you think? -- Markus Lanthaler @markuslanthaler
Received on Thursday, 3 October 2013 20:40:29 UTC