You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-dev@axis.apache.org by Glen Daniels <gd...@sonicsoftware.com> on 2004/07/30 17:12:26 UTC

INVITE: Axis 2.0 summit in Sri Lanka (August 21-24)

Hi folks!

Axis 1.2 is almost out the door, and a variety of discussions have been
simmering for the past year or more about how to improve the package,
restructuring, etc.  As such, it's about time to get some people
together and talk about it in detail.

Sanjiva Weerawarana has managed to get some funding to have the LSF guys
do a lot of work on Axis over the next year, so they would like to host
a meeting in a few weeks in Sri Lanka in order to meet face-to-face and
discuss the design and architecture and logistics for Axis 2.0.  (my
best guess right now, incidentally, is that we'll put out a 1.3 this
year, during which we'll fix bugs and deprecate whatever APIs we think
will be undergoing serious revision for 2.0, and then try to release 2.0
within about 12 months)

Dates are August 21-24 (working over a weekend, as this was the only
time dims could do it), and I'm planning to stay on for a little sight
seeing a few days after that.  Confirmed attendees so far are Sanjiva,
the LSF guys (Srinath, etc), dims, and myself.

Some stuff we'll likely cover:

* Extension architecture including WSDL reading/writing plugins, and a
jar model for encapsulating stuff like WS-* in extension "modules"

* Performance issues, encoding redesign

* XML/Java data binding (JaxME/JAXB/XMLBeans)

* WSDL processing ground-up rebuild

* The build/test process, how can it be improved/streamlined

* etc.

If you're a committer or other party with good knowledge of the Axis
source, please consider yourself invited.  Sanjiva can help out with
local accommodations and travel suggestions.  If you are interested,
please let us know!

For those that can't make it, we can set up an IRC channel and take
minutes.

--Glen

Client Handler bug?

Posted by Shantanu Sen <ss...@pacbell.net>.
I tried to use client side JAX-RPC handlers and saw
the following problem.

I had 2 handlers handler1 and handler 2. If I use the
happy path everything is fine - handler1.handleRequest
is first called and then handler2.handleRequest and
then the corresponding handleResponse methods in
reverse order after invoking the end-point.

However, if I return false in one of the handleRequest
methods, the call still goes through to the service
i.e. the happy path is pursued regardless of the fact
that the client handler request method returned false
signaling the runtime to stop processing.

If I throw a SOAPFaultException or JAXRPCException
from the <handler1/2>handleRequest, the call does not
go through, but I do not see the handleFault methods
called at all.

Is this a known issue? A search on JIRA did not fetch
any bugs related to client side JAX-RPC handlers. 

Looking at the code I see that the AxisClient has no
provisions for handling fault/exception after calling
the JAX-RPC handlers. Here is the code snippet from
AxisClient.invokeJAXRPCHandlers from the latest CVS
tree. 

---------
HandlerChainImpl impl = (HandlerChainImpl)
hiChainFactory.createHandlerChain();

if(!context.getPastPivot()) {
  impl.handleRequest(context);
}else {
  impl.handleResponse(context);
}
impl.destroy();
-----------

What happens if the impl.handleRequest() call returns
false? No action is taken.

On the other hand, for server side JAX-RPC handlers,
when I see the calls from SOAPService, it is indeed
taking care of the false return (I am not sure if it
handles exceptions well - I will test it)
As a result, the server side handlers are working as
expected when a handleRequest returns false.

Here is the code snippet from the SOAPService.invoke

-----------------------------
if (handlerFactory != null)
handlerImpl = (HandlerChainImpl)
handlerFactory.createHandlerChain();
boolean result = true;
try {
  if (handlerImpl != null) {
    try {
     result = handlerImpl.handleRequest(msgContext);
   } catch (SOAPFaultException e) {
     handlerImpl.handleFault(msgContext);
     msgContext.setPastPivot(true);
     return;
  }
}

if (result) {
   try {
     super.invoke(msgContext);
  } catch (AxisFault e) {
      msgContext.setPastPivot(true);
     if (handlerImpl != null) {
      handlerImpl.handleFault(msgContext);
     }
     throw e;
  }
} else {
  msgContext.setPastPivot(true);
}

if ( handlerImpl != null) {
   handlerImpl.handleResponse(msgContext);
}
-------------------------

Am I missing something in the AxisClient which takes
care of these situations? 

Thanks,
Shantanu


Client Handler bug?

Posted by Shantanu Sen <ss...@pacbell.net>.
I tried to use client side JAX-RPC handlers and saw
the following problem.

I had 2 handlers handler1 and handler 2. If I use the
happy path everything is fine - handler1.handleRequest
is first called and then handler2.handleRequest and
then the corresponding handleResponse methods in
reverse order after invoking the end-point.

However, if I return false in one of the handleRequest
methods, the call still goes through to the service
i.e. the happy path is pursued regardless of the fact
that the client handler request method returned false
signaling the runtime to stop processing.

If I throw a SOAPFaultException or JAXRPCException
from the <handler1/2>handleRequest, the call does not
go through, but I do not see the handleFault methods
called at all.

Is this a known issue? A search on JIRA did not fetch
any bugs related to client side JAX-RPC handlers. 

Looking at the code I see that the AxisClient has no
provisions for handling fault/exception after calling
the JAX-RPC handlers. Here is the code snippet from
AxisClient.invokeJAXRPCHandlers from the latest CVS
tree. 

---------
HandlerChainImpl impl = (HandlerChainImpl)
hiChainFactory.createHandlerChain();

if(!context.getPastPivot()) {
  impl.handleRequest(context);
}else {
  impl.handleResponse(context);
}
impl.destroy();
-----------

What happens if the impl.handleRequest() call returns
false? No action is taken.

On the other hand, for server side JAX-RPC handlers,
when I see the calls from SOAPService, it is indeed
taking care of the false return (I am not sure if it
handles exceptions well - I will test it)
As a result, the server side handlers are working as
expected when a handleRequest returns false.

Here is the code snippet from the SOAPService.invoke

-----------------------------
if (handlerFactory != null)
handlerImpl = (HandlerChainImpl)
handlerFactory.createHandlerChain();
boolean result = true;
try {
  if (handlerImpl != null) {
    try {
     result = handlerImpl.handleRequest(msgContext);
   } catch (SOAPFaultException e) {
     handlerImpl.handleFault(msgContext);
     msgContext.setPastPivot(true);
     return;
  }
}

if (result) {
   try {
     super.invoke(msgContext);
  } catch (AxisFault e) {
      msgContext.setPastPivot(true);
     if (handlerImpl != null) {
      handlerImpl.handleFault(msgContext);
     }
     throw e;
  }
} else {
  msgContext.setPastPivot(true);
}

if ( handlerImpl != null) {
   handlerImpl.handleResponse(msgContext);
}
-------------------------

Am I missing something in the AxisClient which takes
care of these situations? 

Thanks,
Shantanu


Re: INVITE: Axis 2.0 summit in Sri Lanka (August 21-24)

Posted by Aleksander Slominski <as...@cs.indiana.edu>.
Glen Daniels wrote:

>
>Some stuff we'll likely cover:
>
i think one of the most important is to redesign message processing 
pipeline to allow true async interactions and to make feasible to 
implement fully WS-RM (here lessons from Sandesha can be evaluated)

i have started few issues in JIRA to keep track of discussion coclusions:
http://nagoya.apache.org/jira/browse/AXIS-1500

>
>* Extension architecture including WSDL reading/writing plugins, and a
>jar model for encapsulating stuff like WS-* in extension "modules"
>
>* Performance issues, 
>
http://issues.apache.org/jira/browse/AXIS-1497

>encoding redesign
>
isnt encoding redesign the same as XML/Java data bindnig redesign?

>
>* XML/Java data binding (JaxME/JAXB/XMLBeans)
>
http://nagoya.apache.org/jira/browse/AXIS-1498

>
>* WSDL processing ground-up rebuild
>
>* The build/test process, how can it be improved/streamlined
>
http://nagoya.apache.org/jira/browse/AXIS-1499

>
>For those that can't make it, we can set up an IRC channel and take
>minutes.
>
good idea and +1 to both.

thanks,

alek

-- 
The best way to predict the future is to invent it - Alan Kay


Re: INVITE: Axis 2.0 summit in Sri Lanka (August 21-24)

Posted by robert burrell donkin <rd...@apache.org>.
On 30 Jul 2004, at 16:12, Glen Daniels wrote:

<snip>

> * XML/Java data binding (JaxME/JAXB/XMLBeans)

it's a bit of a coincidence that this should come up now. i'm messing 
around at the moment with this kind of data binding at the moment. it 
also strikes me that it might be worth adding a betwixt data binder to 
save work having to make the existing bean serializer cope with the 
kind of complex use cases posted recently in JIRA.

i'm subscribed to axis-dev at the moment and reasonably familiar with 
all three projects (JaxMe, XmlBeans and Betwixt) so if any axers 
(what's the right name for a collections of axis committers?) wants to 
start a design thread about the future of data-binding/serialization, 
i'd be enthusiastic.

- robert