On Tue, 2006-03-28 at 19:33 +1200, Dennis Sosnoski wrote: > > +1, which leads me to a quick comment. Currently, the error reporting > > in Axis2 is... less than adequate, to say the least. We're building a > > toolkit/engine which is supposed to make it easy for developers to > > expose and consume arbitrary Web Services in a variety of ways. And > > because there are so many moving parts and extensions, it is EXTRA > > important for us to provide detailed and useful information when > > things go wrong. We can discuss this more on another thread, but I > > think this needs some serious attention if we want good uptake. > > Are you willing to go along on this, Sanjiva? If so, I'll change the > code to print a warning message to the console when a schema with no > targetNamespace is used.
It doesn't work for direct reference from WSDL even for WSDL 1.1 original: part/@type and part/@element both require QNames. The only use is for a schema defining one of those to refer to stuff in a schema with no namespace. Do you know of *any* usecases of this? Can you point me to the WSDL? I'd like to understand it a bit before saying just print a warning. > > -1 > > > > It's great if there's an option to switch BP-compliance-checking on. > > It's not great if there's no way to turn it off. There are plenty of > > people out there who want to be able to use their RPC/encoded > > services, please don't force them to choose another toolkit. Fine Glen .. but someone has to write RPC/Enc support. I have no interest in it (esp. in its full glory; as we've discussed previously the 80% case overlaps with RPC/lit) but if you want to write it that's fine. > I like the BP-compliance-checking switch idea. This would be a lot of > work to implement properly (basically requiring that code be added to > check each recommendation), but would be a major help to users > struggling with interoperability. But it'd be most useful for checking > that the actual runtime service is BP-compliant; checking a WSDL they've > gotten from some other source is probably something that's best done by > a separate tool +1; the runtime should not be doing compliance .. if that's really desired then one can drop a Synapse mediator that does verification. Sanjiva.
