- From: Markus Lanthaler <markus.lanthaler@gmx.net>
- Date: Mon, 3 Mar 2014 14:48:48 +0100
- To: "'Thomas Hoppe'" <thomas.hoppe@n-fuse.de>
- Cc: <public-hydra@w3.org>
+CC public-hydra (keeping Thomas' complete mail) On Monday, March 03, 2014 10:37 AM, Thomas Hoppe wrote: > Hi, > > On 02/28/2014 10:07 PM, Markus Lanthaler wrote: > > On Thursday, February 20, 2014 8:13 PM, Thomas Hoppe wrote: > >> On 02/19/2014 11:16 AM, Ruben Verborgh wrote: > >> Your second mail: > >> > >>> t's still not quite comprehensible > >>> to keep search but drop basic CRUD operations > >>> I'd say: precisely because search is_not_ a basic operation. > >>> > >> Well but then I have to ask why we remove basic stuff like CRUD > >> opertions and keep complex stuff like search... this is somewhat > >> inconsistent to me. > > I do think CRUD functionality is quite important for a lot of APIs. > Over the > > last months, however, it became clear that they confused a lot of > people. I > > would thus be very open to create a separate Hydra CRUD vocabulary > defining > > those operations with strictly CRUD apps in mind. > > ok, I like that idea! > I suggest this as a staring point: > > GET RetrieveResourceOperation > POST AppendResourceOperation (ID is generated and not provided in > the POSTed data) > PUT CreateOrReplaceOperation (ID is provided within the PUTed > data) > PATCH UpdateResourceOperation (Single or multiple property updates) > DELETE DeleteResourceOperation > > The given HTTP actions are not a strict mapping but merely a "best fit" > suggestion > how they could map to the operations. I've amended ISSUE-20 with your proposal: https://github.com/HydraCG/Specifications/issues/20 > There is also a corresponding discussion on SO [1]. > The operations itself should be self explaining. Hmm.. PATCH is always tricky as we would need a separate patch format. > > If we look at collections from that point of view, strictly speaking > > Collections wouldn't belong to the core vocabulary either. The thing > is > > though that they are so common in almost all Web APIs that it does > make > > sense to include them directly in the core vocab... and as soon you > have > > collections, people will ask for search. Now we could move all of > this into > > a Hydra collection vocabulary but I think that would go too far. > > ok, but at least you see my concerns regarding consistency :) > I also think it is no good to split things up in too many parts right > from the start. Exactly > >>> It's not covered by the HTTP spec, > >>> so if machines want to use it, it needs to be described. > >> Yea but as we found out in discussions about the CRUD operations > >> things are more complex than they seem and and thus also need to be > >> described. > >> I'm curious how we will end up with this. > >> > >> To whom it may concern: I will be one week on vacation from now. > > I hope you had some nice holidays. > > Yea, Tuscany is a great getaway :) > > [1] > http://stackoverflow.com/questions/8132445/how-to-support-partial- > updates-patch-in-rest -- Markus Lanthaler @markuslanthaler
Received on Monday, 3 March 2014 13:49:22 UTC