- From: Seaborne, Andy <andy.seaborne@hp.com>
- Date: Mon, 21 Feb 2005 13:22:41 +0000
- To: Dan Connolly <connolly@w3.org>
- CC: RDF Data Access Working Group <public-rdf-dawg@w3.org>
Dan Connolly wrote: > On Mon, 2005-02-14 at 08:52 -0600, Dan Connolly wrote: > >>re ACTION: DanC to Write up definitive service concept >> >>Darn... I thought it was written up in >> http://www.w3.org/2000/10/swap/doc/Reach >>but on review, I see it isn't. > > > OK, somebody reminded me of the tests EricP and > I worked on a while ago... > http://www.w3.org/2000/10/swap/test/dbork/ > > but we don't have a write-up, so here goes... > > In OrderTracking1.n3 , we see... > > <mysql://rdftest@swada.w3.org/OrderTracking/> > is log:definitiveService of > Orders:id, Orders:customer, Orders:product, > Orders:orderDate ... > > The idea is that if you've got a query ala > > SELECT ?productName, ?first, ?last > WHERE > (?o Orders:customer ?c) > (?o Orders:product ?p) > (?o Orders:orderDate ?d) > > (?p Products:name :productName) > > (?c Customers:familyName ?first) > (?c Customers:givenName ?last) > > Then, since each of the properties in the query > is connected to <mysql://rdftest@swada.w3.org/OrderTracking/> > by log:definitiveService , you can delegate the query there; > i.e. if there are any answers to be found, that service has > them. It's definitive. > > I expect lots of SPARQL services are interfaces to SQL > databases, where the data in that SQL database is definitive. > It's the customer database or whatever. All of widgetCo's > customers are in there; they have the power to close that > part of the world. Minor but there isn't often one such database in a WidgetCo. A major source of IT projects is just such data integration exercises. Even small companies have maultiple places - e.g. a sales person's list of customers and the order department. What they need is a semantic web ... > Or it's a database of my CDs. I'm > the authority on which CDs are in my collection, so I > can reasonably publish a description of, say, my desktop > SPARQL query service using log:definitiveService . > > > Other examples in that /dbork/ directory demonstrate federated > query, where cwm will look at a query, realize that it > can be split into parts that can be handled by remote services, > delegate to them, and collect the results. I think our > implementation used the mysql protocol as a network protocol. > That's obviously to be replaced by the SPARQL protocol. > > So that's the idea, on the front side of one page. Does > it merit more investigation? > Yes - :definitiveService seems to be to be a very important piece of information to support federated query and query routing. I don't see that it as an absolute requirment for the WG's service description at this stage though - if it is ready, then yes, put it in; if it feels it needs some refinement by usage then I would not see the LC process being held up for it. Andy
Received on Monday, 21 February 2005 13:24:22 UTC