axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sanjiva Weerawarana" <sanj...@watson.ibm.com>
Subject Re: WsdlAttributes patch
Date Fri, 05 Oct 2001 04:37:42 GMT
> So my question is, should we be generating bindings for one-way and
> notification operations?  To do them properly, I believe the runtime would
> have to change.  No output doesn't just mean a void return, it means NO
> return; make the call and return without waiting for the call itself to
> return.  No input doesn't just mean an empty parameter list, but I'm not
> really sure what it means.

Input-only operations are what you said- you send the output and that's
it .. no waiting for response of any kind because there will not be
one. Output-only means that the service will generate a message of that
type (an event, really) and send it somewhere. Where will it go? That's
not really fully specified within WSDL; that's more specified in a
process flow setting like WSFL where you connect one service to
another. If you connect an output-only operation to an input-only
operation, now you have a proper link.

This is one of the first things we'd like to clarify once standardization
of WSDL begins ..

> The easy thing to do and, barring any 'bugs' like you found, we probably
> already do this by accident, is to treat one-way and notification
> operations as request-response operations.  But is it the right thing to
> do?  Personally, I lean toward "yes" since most folks won't understand
this
> little nuance and would be rather irate if Wsdl2java refused their WSDL.
> If we DO refuse it, we should give them a meaningful reason why.

I think that given that the semantics are fundamentally different, the
proper thing to do is not support input-only, output-only and solicit-
response operations at this time.

Sanjiva.



Mime
View raw message