- From: Olaf Hartig <hartig@informatik.hu-berlin.de>
- Date: Mon, 1 Aug 2011 21:30:04 +0200
- To: public-prov-wg@w3.org
Hello, On Friday 29 July 2011 08:44:14 Graham Klyne wrote: > OK, I'll work on something like that and transfer and adapt the SPARQL > stuff to section 4 (provenance querying). I wouldn't remove that from Section 3.4. Instead, we should understand it as an alternative to a REST based API. However, the issue for a client then might be the following: Given a resource/document URI u_1 and the URI u_2 for a third-party provenance registration service; we want to ask that service for a provenance URI u_3 that refers to provenance information about u_1. How do we decide whether u_2 has to be accessed as if it were a SPARQL endpoint or as if it were the REST API? > I'll assume that only the simplest discovery case is needed; i.e. have URI > of entity, what to get URI(s) for its provenance. > > (The other case that used other forms of identification to locate > provenance will get dropped from 3.4, since that was really just showing > how to exploit SPARQL capabilities) That's exactly what I liked about the SPARQL based mechanism. The provider of a third-party provenance registry service may support such, more complex search patterns and, thus, set itself apart from other registries that only support simple queries with a single prov:hasProvenance triple pattern (or the REST API equivalent thereof). It's a business opportunity. Cheers, Olaf > > #g > -- > > Reza B'far wrote: > > As an implementer, I strongly agree with Luc. REST-base http protocol > > with minimal payload is a much more likely first option. Also, it's not > > necessary that provenance is stored in a traditional notion of a > > database. I guess folks call many various persistence mechanisms these > > days a database, but some of these (say simple files, LDAP, etc) map mug > > easier to a simple API with just GET's or something like that. > > > > Usage of SPARQL is still very limited to sem web specialists (compared to > > simple REST/http/minimal) > > > > Best > > > > On Jul 28, 2011, at 8:04 AM, Graham Klyne <GK@ninebynine.org> wrote: > >> SPARQL might well be considered overkill if you had to implement it from > >> scratch. But there are now many SPARQL toolkits that doesn't seem to > >> be such an issue. If you do a REST interface, it would likely be > >> backed by some kind of database, so not so much complexity saving > >> there. > >> > >> I think there's a case for defining a simple REST API that might be > >> implemented over a SPARQL store using an LDA > >> (http://code.google.com/p/linked-data-api/wiki/Specification) toolkit, > >> or by other means. It's not clear to me if that would receive more > >> developer attention. Maybe for the time being we might define both > >> (REST and SPARQL) and review them as we approach candidate > >> recommendation. (Oh, but that won't apply to the PAQ, will it?) > >> > >> (BTW, SPARQL is not inherently non-RESTful for simple queries.) > >> > >> #g > >> -- > >> > >> Provenance Working Group Issue Tracker wrote: > >>> PROV-ISSUE-53 (sparql-query-is-overkill): can't we have a lighter > >>> method to retrieve provenance-uri, given a document uri? [Accessing > >>> and Querying Provenance] http://www.w3.org/2011/prov/track/issues/53 > >>> Raised by: Luc Moreau > >>> On product: Accessing and Querying Provenance > >>> Isn't it too onerous to require third parties to provide sparql > >>> endpoints, simply to answer the query illustrated in > >>> http://dvcs.w3.org/hg/prov/raw-file/default/paq/provenance-access.html > >>> #third-party-services (give the provenance-uris for this resource)? > >>> Stian had suggested a REST service to answer this query, encoding the > >>> URL as an argument passed to the service. I had outlined a variant > >>> with GET instead of POST. Some implementers may not use an RDF > >>> serialization of provenance, and therefore, it would be good to have a > >>> non-SPARQL approach for finding a provenance URI.
Received on Monday, 1 August 2011 19:30:40 UTC