We're not going to ditch SOAP because of the advantages of code generation from WSDL (top-down methodology), WS-* standards and features, etc. As long as our code only use JAX-WS APIs, we're not tied to a specific implementation.
RESTful invocation is a minor use-case. The HTTP headers idea is an interesting idea, however, not workable if we want to invoke with a browser. (although the browser invocation use-case is just for testing, right now) Thanks, -Chris W. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Brad Sent: Friday, August 01, 2008 4:03 AM To: [email protected] Subject: Re: ArrayIndexOutOfRange exception problem in URIMappingInterceptor Hi, if you want to REST and a plain POST of an xml document then why not ditch SOAP altogether and use HTTP headers for your metadata? Both of your requirements are essentially XML over HTTP so why not use their common ground to your advantage. The schema part of your WSDL will still be valid so your interface "contract" is still there. Saves you have to do too much which ties your code to a specific implementation too. Brad. On Fri, Aug 1, 2008 at 3:05 AM, Wolf, Chris (IT) <[EMAIL PROTECTED]> wrote: > We want to deploy services that are consumable either via conventional > POST of a request document, or via RESTful invocation. It seems that > wrapped-doc/literal has the feature of being able to do that without > resorting to any extra work via the URIMappingInterceptor, which is > wired in by default. > > The trouble is that our request has some meta data that must be added > to the soap header, and in the case of conventional invocation via > POST of a request document, all is fine. > > Now we want to also be able to pass this meta data when performing > RESTful invocation, so the idea is to just tack on some extra query > string parameters, in addition to the operation parameters defined in > the WSDL. > > Ideally, the RESTful invocation processor would ignore the extra > parameters which are not in the WSDL, unfortunately, before > URIMappingInterceptor even gets a chance to complain, we get an array > index error. > > > protected MessageContentsList getParameters(Message message, > BindingOperationInfo operation) { > [...] > int idx = 0; > if (inf != null) { > idx = inf.getIndex(); > } > Class<?> type = types[idx]; <---- line #178, CXF-2.1.1 > (types is null for no-arg operation) > [...] > } > > I could subclass this interceptor and change the behavior of > getParameters(...) by override, > then somehow configure cxf to use the derived class. However, I > thought I'd ask the list - is this idea of simulating the Envelope > Header in RESTful invocation by exra parameters a good idea? > Obviously, there are limitations to what can be represented compared > to full hierarchical XML, but this can be partially compensated for by > a param name scheme such as: > group0.item0=foo&group0.item1=bar&group1.item0=yin&group1.item1=yang > > I would hate to have to muddy up our contract definitions in the WSDL > just to accommodate meta data that is only relevant to the > infrastructure, rather then specific business functionality. I ask > beause it might just be easier for me to patch URIMappingInterceptor > directly, but I wonder if the patch would be accepted. > > Thanks, > > -Chris W. > -------------------------------------------------------- > > NOTICE: If received in error, please destroy and notify sender. Sender does not intend to waive confidentiality or privilege. Use of this email is prohibited when received in error. > -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not intend to waive confidentiality or privilege. Use of this email is prohibited when received in error.
