axis-c-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Samisa Abeysinghe <samisa_abeysin...@yahoo.com>
Subject Re: Why create new client transport object for each invoke?
Date Tue, 09 Nov 2004 09:44:56 GMT
Hi Damitha,
   Good news: it works fine with HTTP/1.1 and it is not slow with Axis C++ server. (I had
made a
mistake earlier)
   
   TODOs: Need to get the same working for HTTP/1.0

   I have commited the solution to the CVS. Please have a look at the CVS mail message for
more
details.
   Please let me know if you run into problems.
Thanks,
Samisa...



--- Roshan Weerasuriya <roshan@opensource.lk> wrote:

> hi,
> 
> I can't see a reason why the server is becoming slow due to this. But as
> you say so if it is , it is intresting to investigate it.
> 
> Roshan
> 
> 
> On Tue, 2004-11-09 at 14:12, damitha@opensource.lk wrote:
> > Hi Samisa,
> > Yes it could affect the server. Please tell how you manage to do it?
> > thanks
> > damitha
> > > I managed to get it working without closing connection all the time. But
> > > the client runs very slow
> > > with Axis C++ server.
> > > With Axis Java the speed is OK.
> > > I wonder if this is something wrong on server side.
> > >
> > > Samisa...
> > >
> > > --- damitha@opensource.lk wrote:
> > >
> > >> Hi Samisa,
> > >> > The current client side code creates a new transport for each method
> > >> > invocation from within the
> > >> > same Stub, no matter we use HTTP 1.1 or 1.0. With HTTP 1.1, this is
> > >> > incorrect, as by default, it
> > >> > should assume persistant connections. When implementing Keep-Alive
> > >> > support, it is important that
> > >> > the client transport follow the given HTTP rules.
> > >> >
> > >> > I could put some effort and fix this (and I think we should fix this).
> > >> > However, I want to first
> > >> > make sure that I am not breaking any serious assumptions made by the
> > >> code.
> > >> > So if any of you have
> > >> > any reasoning for creating the transport object for each invocation,
> > >> > please let me know.
> > >>
> > >> I cannot remember any aparent reason why this is so. Anyway I tried to
> > >> introduce an initialize method in transport and then create the
> > >> transport
> > >> object while in the Call constructor so that transport is created once.
> > >> Then in each call initialize it will call transport's initialize method
> > >> without recreateing it. Then I found that in the Stub with several
> > >> methods, only the  first method get executed successfully and successive
> > >> method calls result in segfault. I need to have a more close look into
> > >> this.
> > >>
> > >> thanks
> > >> damitha
> > >>
> > >>
> > >> >
> > >> > Thanks,
> > >> > Samisa...
> > >> >
> > >> >
> > >> >
> > >> >
> > >> > __________________________________
> > >> > Do you Yahoo!?
> > >> > Check out the new Yahoo! Front Page.
> > >> > www.yahoo.com
> > >> >
> > >> >
> > >> >
> > >> >
> > >>
> > >>
> > >
> > >
> > >
> > >
> > > __________________________________
> > > Do you Yahoo!?
> > > Check out the new Yahoo! Front Page.
> > > www.yahoo.com
> > >
> > >
> > >
> > >
> > 
> > 
> 
> 



		
__________________________________ 
Do you Yahoo!? 
Check out the new Yahoo! Front Page. 
www.yahoo.com 
 


Mime
View raw message