- From: Alexandre Bertails <alexandre@bertails.org>
- Date: Wed, 30 Jul 2014 17:03:48 -0400
- To: Steve Speicher <sspeiche@gmail.com>
- Cc: "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>
On Wed, Jul 30, 2014 at 4:54 PM, Steve Speicher <sspeiche@gmail.com> wrote: > On Mon, Jul 28, 2014 at 1:18 PM, Alexandre Bertails <alexandre@bertails.org> > wrote: >> >> All, >> >> The last thread on the interaction model reminded me of a question we >> have at work. >> >> At Pellucid, we retrieve data from containers using a custom language. >> For now, it does filtering, ordering, inlining, and grouping. No >> SPARQL. No paging at the HTTP level like in LDP. It's RDF data all the >> way. >> >> Here is the way that works at the protocol level: the container >> _itself_ supports the filtering API, and we POST the query to it. As >> you can see, we violate the semantics of an LDPC, as we are not adding >> the query to the container. We don't really want to have a separate >> endpoint for each container. We would be happy to use HTTP SEARCH or >> QUERY, if they were available or not overloaded. > > Alexandrei (aka Alexandre and Andrei), > > Instead of a new HTTP verb, would using POST + query parameter be worse: > POST <container URL>?<search param> ? I am not sure it would make a big difference: if an empty document is a valid Turtle document, then I guess this could create an empty graph. Also, there is always the issue of reaching the size limit for the query string. > I understand that this another resource on the web, different URI, but would > be the same endpoint (ie entry point in your server impl). You'd need a way > anyway for your "LDP on steroids" client to know it can safely send these to > execute a query vs create a resource with this content. In other words, I > would think you'd want a discover mechanism wish would expose a URI to the > endpoint for search, which could just be getting back some header values or > data from the container itself, URL-template, etc. > > Curious to hear how you do discover and sharing of the endpoint. No special strategy. We just assume the LDPC supports the added protocol, as an extension. Hence my questions on how to do it right. We don't rule out the possibility to Link another resource in the future. Alexandre > > Cheers, > Steve > >> >> >> Is there a good way to do that while being fully compliant with the >> spec, *and* ? This is the same question for people who wants to add >> SPARQL query capabilities to an LDPC. >> >> Alexandre >> >
Received on Wednesday, 30 July 2014 21:04:15 UTC