I've update the jaxws frontend so that Metadata/@wsdlLocation attribute contains a pair of values (targetNS plus the actual location) by default. If needed this can be 'reverted' by setting an "org.apache.cxf.wsa.metadata.wsdlLocationOnly" contextual property on a given endpoint.
As a side note, I moved the NO-OP LocatorServiceImpl from testutils closer to the actual Locator test and used this impl to test the changes rather than introducing some more implementations.... cheers, Sergey On Tue, Aug 31, 2010 at 4:39 PM, Sergey Beryozkin <[email protected]>wrote: > Hi, > > I've had an interesting chat with Richard Opalka today regarding the way > CXF builds the Metadata part of the (WSA) EPR. Particularly, at the moment, > CXF does something like this : > > <Metadata wsdli:wsdlLocation='http://localhost:8080/hello?wsdl' > xmlns:wsdli='http://www.w3.org/ns/wsdl-instance'>...</Metadata> > > Richard pointed me to the relevant section of the WSDL 2.0 spec which also > mentions WSDL 1.1. According to that section, @wsdlLocation is expected to > contain pairs similarly to the way an xsi:schemaLocation attribute does, ex, > > wsdli:wsdlLocation='http://helloservice <http://localhost:8080/hello?wsdl> > http://localhost:8080/hello?wsdl' > > where "http://helloservice"; is the target namespace of the referenced WSDL > instance. > > I'm not understanding where this requirement to have the pairs come from > originally (i.e, a given service endpoint is described by a single WSDL doc > (which may import few others) as opposed to a xml instance validation > depending on a number of schemas being available) and at least on the > surface the first value is redundant, but it needs to be fixed anyway > because there's a "MUST" bit in [1]. I would not even worry if it were > "SHOULD". > > > My question is : can it be that CXF omits the target namespace (such as > http://helloservice) for a reason ? Example, can we have legacy consumers > parsing a wsdlLocation attribute manually and assuming its value is a > location only as opposed to a pair of targetNs/location ? I don't recall now > but I vaguely remember I was doing some manual parsing of the returned EPRs > awhile back... > > Do any users out there have the applications depending on this attribute ? > > I'm just wondering how to fix it. I guess the contextual property can be > introduced if there's a concern that fixing it can lead to > backward-compatibility issues... > > cheers, Sergey > > [1] http://www.w3.org/TR/2007/REC-wsdl20-20070626/#wsdlLocation-aii >
