axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chathura Herath" <>
Subject Re: [Axis2] WSDLServiceEmitter
Date Mon, 13 Dec 2004 05:13:21 GMT
On Sun, 12 Dec 2004 14:52:36 -0500, Glen Daniels wrote
> Sanjiva Weerawarana wrote:
> > Damn, I feel bad for Glen and Dims .. they have a lot of email
> > waiting for them when they wake up :-).
> !!!! No kidding. Took me a couple days to get to it. :)
> >>There is nothing wrong in creating a new document structure because the
> >>performance and memory is not an issue in this case, because this will be
> >>done offline, not at runtime.
> I don't agree with this.  What about the case where we dynamically 
> generate WSDL from deployed Java services?  That's a runtime issue. 
> Plus, various deployed Handlers/Modules might need to affect the 
> WSDL generation as well, which also pulls this into the runtime realm.

Think you got a point.

> > Well you should be able to go from WOM to WSDL 2.0 without any
> > problem (or the WSDL WG has screwed up!). However, I do understand
> > that its convenient to have a tree model to generate it rather
> > than having to do a one-pass stream out to a StAX push writer.
> Why?  As I see it you've already GOT the model (the WOM itself), so 
> why not just have the emitter write that directly via the push API?  
> I don't understand why that should be difficult, since you already 
> know in what order all the dependencies have to be written, etc....

Direct serialisation of WOm looks fine to me when considering the issues 
raised. My only concern is that we should keep the Serialization not burnt 
into WOM itself. Reason why i say that is there should be room for different 
serializations, which i belive is something that has been upheld by the 
concept of Component model in WSDL 2.0. 

> Despite very much believing in "eating our own dog food", I think 
> you need to know when you need to eat at all. :)  I'm +1 to 
> serializing WOM directly via the push API without an intermediate OM 
> model if at all possible.
> --Glen


Lanka Software Foundation (

View raw message