- From: Richard Cyganiak <richard@cyganiak.de>
- Date: Thu, 18 Sep 2008 12:29:21 +0100
- To: Bob DuCharme <bob@snee.com>
- Cc: public-lod@w3.org
On 18 Sep 2008, at 02:48, Bob DuCharme wrote: > As a side note, I think this is going to be very big, because while > Linked Data (and much of the semantic web) is theoretically about > exposing data to programs instead of to eyeballs like the > traditional web, most of the linked data and semantic web demos I > see out there are about visual browsing of linked data--displaying > it to eyeballs. When we can grab the results of a linked data SPARQL > query with a script, then we can really start doing new and > interesting things with it. I agree. Let me add some random thoughts on how to take this idea (retrieving data from RDF sources on the command line for further processing) into more powerful directions. 1. SPARQL is great, but too verbose for the command line. I would really like to use a single-line, XPath-style query language for RDF in situations like this. There are a couple of proposals for such a language out there, several of them are called "RDFPath" and there's Uche Ogbuji's Versa. It would be really great to see some progress in this area. The first one of these languages to get a well-packaged implementation on top of the Jena API will probably win. 2. I really like the idea behind the Semantic Web Client Library (part of NG4J) for exactly the reason you state above -- it makes it possible to get data from the sources out there for further processing. It has a command line tool that allows you to ask SPARQL queries against (potentially) the entire Web. Unfortunately, we never managed to get the library beyond a prototype stage, and a lot of work remains to be done on it. Still, the idea remains valid and is important, I think. 3. We are all waiting for SPARQL processors that federate multiple SPARQL endpoints transparently into a single endpoint. Progress is being made in this area, but it's slow. Meanwhile, there is a very nice 80/20 solution to this problem: Andy Seaborne has implemented a SERVICE keyword for his extended variant of SPARQL, which allows you to address parts of a SPARQL query to a specific endpoint. This seems like an easy win for data integration demos. Richard > > > Bob
Received on Thursday, 18 September 2008 11:30:01 UTC