- From: Thomas Hoppe <thomas.hoppe@n-fuse.de>
- Date: Wed, 25 Feb 2015 00:44:06 +0100
- To: public-hydra@w3.org
Hi, I was following the recent discussions and I agree with what you said, Markus, I want to add my summary: 1.) We should be able to indirectly express that a collection in terms of hydra whose members are in a foaf:knows relationship with some subject, actually means that this relationship is true for every member of the collection. As far as I have understood, the `hydra:manages` construct allows just that. And I think this is not only useful for the often stressed reasoner; This info can also be leveraged by a client that happens to understand the construct in hydra and can react upon it... so no AI or reasoning required here. 2.) I'm against having the members of the collection outside of the collection. 3.) I understand people who argue that it would be nice to _also_ express paging in terms of a offset/ limit tuple. Yes, this is not good for those "streamy resources" like newsfeeds but for ranged resources it is a way to express this and it is heavily used in the wild. So we would help people to describe existing APIs if we can offer this. I don't ask for adding this to the vocab but we could offer some advice (also outside of the spec) how to model such a construct, maybe with hydra's `firstPage`, `lastPage`, `nextPage`, `previousPage` and IRI templates as proposed by Greg. 4.) I think Ruben is right and we should have the concept of singe pages/ partial collections/ whatever because the alternative just seems to just wrong. Greets, Thomas On 02/23/2015 10:48 PM, Markus Lanthaler wrote: > I think we are starting to go in circles regarding ISSUE-42. Would everyone > agree with the following statements? Is there anything I forgot? > > > In Web APIs, we often need to return a lot of related data, e.g., persons > someone knows > > Sometimes it is too much data to be returned in a single response > > Therefore, we would like to split this data and return it in multiple > responses instead > > Nevertheless, the client needs to be able understand, that all these > responses are actually just partial views of a big "collection" (the persons > someone knows) > > Due to the way some vocabularies are defined, we can't link directly to such > "helper resources" as that would be misinterpreted by clients (a client > would misinterpret a helper resource to be a person if it would be linked to > via foaf:knows e.g.) > > We want the relationships to be explicitly expressed so that we don't have > to rely on a reasoner > > If JSON-LD is used as the serialization format, the documents should look as > idiomatic as possible. I.e., they should closely resemble current Web APIs. > > > > -- > Markus Lanthaler > @markuslanthaler > > > >
Received on Tuesday, 24 February 2015 23:44:35 UTC