- From: Andrew Hacking <ahacking@gmail.com>
- Date: Wed, 25 Feb 2015 20:17:42 +1000
- To: Dietrich Schulten <ds@escalon.de>
- Cc: Markus Lanthaler <markus.lanthaler@gmx.net>, public-hydra@w3.org
Received on Wednesday, 25 February 2015 10:18:15 UTC
On 25 Feb 2015 17:48, "Dietrich Schulten" <ds@escalon.de> wrote: > In our case the attribute of a json-ld is the rel, the @id is the href. The rel says that the current document is related to another one in a well-defined way (e.g. next, prev, create-form [1], foaf:knows). > > As it stands right now, that works very elegantly with PUT, PATCH or DELETE as long as a single item is updated or deleted. > > However, when it comes to creating new items with POST or PUT, it turns out that we can't use the @id as elegantly. > > The GET aspect made us go in circles already :-) > > Maybe that sheds some more light to the discussion, or it makes more clear where I come from. You raise practical considerations. In my implementations I prefer to create resources using PUT with client assigned identifiers. These PUTs often include within the payload other resources that must be created or updated as a logical idempotent transaction. The resource being PUT represents an aggregation of resources, and the URL reflects that eg /invoices/123?includes=items which is distinct from /invoices/123 that does not include line items. This is the auxiliary resource capability I have raised before and applies equally to GET/PUT/PATCH/DELETE for both transactional changes and avoiding 1+N queries. It may have some overlap with collections/pages including related resources for efficiency vs just being a descriptor.
Received on Wednesday, 25 February 2015 10:18:15 UTC