I would prefer that whatever compiles the jws also does semantic validation of the jws. So, if you drop a jws file into axis, an error related to the use of the annotations (e.g., WebResult & Oneway) would be treated in the same way as a syntax error in the file (e.g., missing semi-colon). I do not think these semantic checks should be implemented in the deployment code. Rahter, they should be implemented in the apt/mirror AnnotationProcessor, and that should be run when compiling a jws. If you implement the checks to run on an already-compiled .class file, then you'll have two different sets of checking logic - one in the deployment code, and another in the apt/mirror AnnotationProcessor.
All that said, I agree there should be some deployment-time/run-time validation of artifacts to ensure they are consistent, but I would expect these errors to be pathalogical (i.e., any user-introduced error should have been detected by the annotaiton processor at compile-time). My .02, Mike > -----Original Message----- > From: Daryoush Mehrtash > Sent: Tuesday, September 21, 2004 11:55 AM > To: Beehive Developers > Subject: RE: Oneway and WebResult > > > The IDE that you use may or may not be 181 aware. > > You are correct, if you use an IDE that is 181 aware then the > compile-time semantic validation process will flag the errors. > > My question is more concerning the deployment phase. If you > drop a file > that didn't pass the semantic check (at least in this case) should the > server make some assumption (e.g oneway takes precedence over > webresult), print warning messages, and just go ahead with > deployment, > or throw and exception an fail the deployment all together. > > daryoush > > -----Original Message----- > From: Mike Kaufman > Sent: Tuesday, September 21, 2004 9:05 AM > To: Beehive Developers > Subject: RE: Oneway and WebResult > > > I don't understand how it is possible for a jws to able to be > deployed, > but hasn't been semantically validated. Isn't it the case that > "processing" a jws file enforces semantic rules as well as > generates the > artifacts necessary to deploy & run? > > That said, the mutual exclusivity of oneway & webresult should be an > error in the compile-time semantic validation. As a sanity > check, if a > jws ever comes through that includes both oneway & webresult, there > should be an exception, either during deployment, or when the > erroneous > method is executed. > > Mike > > > > -----Original Message----- > > From: Daryoush Mehrtash > > Sent: Monday, September 20, 2004 7:41 PM > > To: Beehive Developers > > Subject: RE: Oneway and WebResult > > > > > > > > What should the correct behavior be? > > > > If a method has @WebResult and @OneWay we can either: > > > > a) ignore webresult > > b) ignore oneway > > c) throw an exception > > > > > > > > There are two use cases to consider: > > 1) IDE > > 2) Server > > > > In the IDE case validation process (TBD) can report the error. > > > > In the Server case (e.g. .jws that is deployed to the axis) > we need to > > consider the case were the .jws file was developed in an editor that > > didn't enforce the 181 semantics. Should we refuse to deploy > > the file, > > or just give a warning message and ignore either webresult, > or oneway? > > > > Daryoush > > > > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > > Sent: Monday, September 20, 2004 6:25 PM > > To: Beehive Developers > > Subject: Re: Oneway and WebResult > > > > Hello Jonathan, David. > > > > Thank you for responding to my silly questions everytime. > > > > I wanna make sure one more thing. > > The WebResult and Oneway annotations can appear on the same method ? > > I mean no exceptions will be thrown. > > (The current codes seem to allow it. > WebServiceMETHODMetadata doesn't > > handle > > that kinda situation.) > > > > Like your previous mails , even though the WebResult > > annotation appears, > > > > the WebResult annotation will be ignored if the Oneway > annotation also > > appears > > on the same method. > > > > Is it true ? > > > > Thanks in advance. > > > > wolfgang > > > > > > > > > > > > > > >
