On Jan 26, 2017 8:21 AM, "Ruben Verborgh" <Ruben.Verborgh@ugent.be> wrote: Hi Albert, > No more hard-coded queries in your Linked Data consuming applications! Why not? Isn't the whole point of the Semantic Web that clients should know what they want, instead of servers needing to know what clients want? Or to say it differently: the more specific an API, the less clients that are able to use it? No, or to say it differently, yes, or to say it less differently, yes and no, prime minister :-) The main point of the semantic web is that clients should know[1] what they got. To the extent that the domain specific view may be used in contexts where there are stronger shared Ontological commitments, it would seem that such a view is not in conflict with semantic web principles. Further, to the extent that the result is generated by a complex query requiring extensive knowledge of the internal organization and vocabulary of the provider's knowledge base, as well as the performance characteristics of the backing implementation, the simplified API may be preferable. Finally, SPARQL is not a simple query language to learn, and it can be much too easy to copy and paste one's way to a cluster melting Query Of Doom. One rarely finds open SQL access as part of an API, especially on purpose... On the other hand, exposing a raw SPARQL endpoint will allow for richer queries, and it may be easier to add such an endpoint to a federation that does not support real time nose following. I'm not sure if there are LOD SPARQL endpoints that allow UPDATE ; that'd be a different kettle of kettles. Simon [1] Where "know" means that the to the extent that the client and server share vocabulary, the client's "beliefs" about the "beliefs" that the server "intended" to convey are in accord.
Received on Thursday, 26 January 2017 16:47:42 UTC