I understand, that's why I am looking for just a hint about the architecture - can somebody tell me which CXF class or which another library is responsible for serializing the SOAP request. If I find a proper point, I can download sources, debug and experiment with it. There must be some difference in IBM java for AIX and other system which definetely cause my problem. If I will know who is responsible for serializing soap request in cxf it will be easy to find which API from jdk is called and findout the difference then maybe proivde some patch...
Dne úterý 25 září 2007 23:44 Benson Margulies napsal(a): > This is going to be mighty difficult to make any sense of. I, for one, have > no access to an environment to repro this in. > > > -----Original Message----- > > From: Bc. Jiří Mikulášek [mailto:[EMAIL PROTECTED] > > Sent: Tuesday, September 25, 2007 3:45 PM > > To: cxf-user@incubator.apache.org > > Subject: Missing prefix in request on IBM > > > > Hello, > > yet another question regarding WS-client development on IBM. > > This issue appears only on IBM java on AIX os. Using IBM java on another > > operating system everything works fine. > > > > There is missing prefix by operation element in SOAP request. We are > > using document/literal wrapped binding > > > > Our request on IBM looks like (without an envelope): > > > > <soap:Body> > > <Vendor xmlns:ns2="some:namespace"> > > .... > > </Vendor> > > </soap:Body> > > > > but should look like (and looks like on windows or linux): > > > > <soap:Body> > > <ns2:Vendor xmlns:ns2="some:namespace"> > > .... > > </ns2:Vendor> > > </soap:Body> > > > > The missing prefix is problem for the service part (not developed by us > > and > > running on webmethods) and the operation couldn't be called - we > > got "Service :Vendor does not exist". > > > > Please can somebody give me a hint which part of cxf or another library > > is responsible for serializing the SOAP request and what jdk dependent > > API is used? I must try to findout the primary cause otherwis this > > problem cannot be > > fixed. Our tests have just proven that the xerces implementation has no > > influence. > > > > Thank you > > > > Jiri Mikulasek
