- From: David Charboneau <dcharbon@us.ibm.com>
- Date: Tue, 5 Jan 2010 09:13:12 -0500
- To: public-rdf-dawg@w3.org
- Message-ID: <OFC5C8C718.DB5F11F7-ON852576A2.004DDE2E-852576A2.004E1D7E@us.ibm.com>
Hi Everyone, I had a few questions for the group regarding updates to the SPARQL Protocol. 1. Should we update the WSDL specification references to the latest recommendation? They are currently referencing a 2006 draft of WSDL 2.0, rather than the current WSDL 2.0 recommendation. 2. Only a POST binding for the update operation is specified in the WSDL since GET should have no side effects. Are there any objections to this? 3. There are two faults specified for graph errors in an update request, GraphDoesNotExist and GraphAlreadyExists. Are there other faults that should be specified? Or, should we even go this fine grained? We could instead go with a more generic UpdateSemanticError or something and let the service implementers decide how to provide clients with more detailed information about the semantic error. This is for the case when an update is specified to a nonexistent graph, or a graph is specified to be created and it already exists. 4. My reading of http://www.w3.org/TR/wsdl20/#meaning is that a service that claims to implement our WSDL for the protocol would have to implement both the query and update operations. We'd prefer that these were optional - a service provider could implement one or both. To that end, we could either define two distinct interfaces; or we could add a new fault that allows a service to report "Operation Not Supported" in order to satisfy the single interface but opt out of implementing one of the operations. Which of these is preferable? Or, is my reading of the WSDL specification incorrect? Thanks, David Charboneau dcharbon@us.ibm.com
Received on Tuesday, 5 January 2010 15:16:57 UTC