Problems with oneway methods

Mark Little M.C.Little@ncl.ac.uk
Wed, 10 Dec 1997 10:06:49 +0000


> > When we create 2 CORBA objects dynamicaly (one client A and one server B) in
> > the same program (program.exe), its seems that the oneway methods of the
> > server do not have the expected behavior (if A calls the oneway method
> > B->service() then it has to wait until the end of the method service()
> >
> > Is there a way to correct this problem ?
> 
> As far as I understand, this is the normal way. You could use a thread
> to submit calls to the server object, in which case the client does not
> have to wait.
> 
> By the way, if the server object method raises an exception, the
> exception will show up in the client's program as an UNKNOWN exception.

The functionality required from one-ways is specified (no guaranteed
delivery, best-effort, etc.) , but their implementation is pretty much
up to the ORB. For example, there are a couple of ORBs which simply
ignore one-ways, and this is seen as conforming to the standard. As
far as using them in a portable manner is concerned: forget it! The
OMG is currently working on a new asynchronous message passing model
which should do in a standard way what most people expect one-ways to
do.

Cheers,

Mark.

-----------------------------------------------------------------------
SENDER : Dr. Mark Little, Arjuna Project, Distributed Systems Research.
PHONE  : +44 191 222 8066, FAX : +44 191 222 8232
POST   : Department of Computing Science, University of Newcastle upon
	 Tyne, UK, NE1 7RU
EMAIL  : M.C.Little@newcastle.ac.uk