With the latest changes done to axis3 transport by Fred, this too seem
to be fixed as well, not at the Call class level but at Transport
level (most probably in ChannelFactory class)
I think it is a better solution.
Thanks,
Samisa...
On Tue, 25 Jan 2005 09:54:36 +0000, John Hawkins <[EMAIL PROTECTED]> wrote:
>
> Sounds like a good idea +1
>
>
>
> John Hawkins
>
>
>
>
>
> Samisa Abeysinghe <[EMAIL PROTECTED]>
>
> 25/01/2005 07:59
>
> Please respond to
> "Apache AXIS C Developers List"
>
>
> To Apache AXIS C Developers List <axis-c-dev@ws.apache.org>
>
> cc
>
> Subject ChannelFactory::LoadChannelLibrary is called for each method call on
> Stub
>
>
>
>
>
> Hi,
> It looks to me that the above mentioned method is called for each
> method invocation.
> This is because we have
> m_pTransport->setTransportProperty( CHANNEL_HTTP_DLL_NAME,
> pcChannelHTTPLibraryPath);
>
> inside Call::openConnection() and this is called by Call::initialize()
> which is always called when a method is invoked.
>
> The code works OK, but the problem is in case of performance.
> 1. A Channel object is created for every method invocation
> 2. Because of 1 above, we always have to create a new connection
> 3. 2 above violates Connection-keep-alive in HTTP/1.1
>
> In axis2 transport, we had the featurer, where the connection to
> service is re-established iff the URL changes, thus re-using the same
> connection for subsequent invocations in case of HTTP/1.1
>
> Because of the channel loading mechanism, it is a but though to do the
> same in case of axis3. But I think if we conditionnaly invoke
> m_pTransport->setTransportProperty( CHANNEL_HTTP_DLL_NAME,
> pcChannelHTTPLibraryPath); if and only if pcChannelHTTPLibraryPath has
> changed, then we could increase the efficancy.
> Using something like:
> if( !strcmp(pcChannelHTTPLibraryPath, pcOldChannelHTTPLibraryPath)
> {
> m_pTransport->setTransportProperty( CHANNEL_HTTP_DLL_NAME,
> pcChannelHTTPLibraryPath);
> }
>
> If we change this, the connection related stuff will work as axis2 as
> we used the same code here in axis3.
> I did not change any code in CVS because I am not sure if the above
> would have any side effects or any other related concerns.
>
> Thoughts please.
>
> Thanks,
> Samisa...
>
>
