[omniORB] corbaloc/corbaname

[Richard Hardgrave]

I believe I have found my misconception from reading the CORBA standard
on corbaloc & corbaname.
The omniORB guide says that the usual IOR string is considered a "URI".
But, it's not the other way around.  You cannot just drop in a corbaname
in where you might use an IOR string.   In some cases you can, but the
object reference you get back has to be listening (or have an agent
listening) on the port named in the URL.
To me, it looks like corbaname might work for eg3_clt if the
object it was looking for was the Naming Service on a remote machine.


> From owner-omniorb-list@uk.research.att.com Wed Feb  6 18:02 CST 2002
> Date: Thu, 07 Feb 2002 00:54:37 +0100
> From: bjorn rohde jensen <shamus@tdcadsl.dk>
> X-Accept-Language: en
> MIME-Version: 1.0
> CC: omniorb-list@uk.research.att.com
> Subject: Re: [omniORB] corbaloc/corbaname
> Content-Transfer-Encoding: 7bit
> X-Loop: omniorb-list@uk.research.att.com
> 
> Hi Richard,
> 
>  You could use a corbaname URI with the eg2 demo this way;
> Start a nameservice on the localhost
> 
> bjorn@twilight ~ $ /usr/local/omni/src/examples/echo/eg2_impl 
> 
> then bind the IOR of the object dumped by eg2_impl in the naming graph
> under fx test/eg2
> 
> bjorn@twilight ~ $ /usr/local/omni/bin/i586_linux_2.0_glibc2.1/nameclt
> bind_new_context test
> 
> bjorn@twilight ~ $ /usr/local/omni/bin/i586_linux_2.0_glibc2.1/nameclt
> bind test/eg2
> IOR:010000000d00000049444c3a4563686f3a312e3000000000010000000000000064000000010102000b0000003137322e32322e302e310000a28000000e000000fe6bbd613c000006f7000000000000000200000000000000080000000100000000545441010000001c00000001000000010001000100000001000105090101000100000009010100
> 
> then run the eg2_clt using the corbaname
> 
> /cpp/corba $ /usr/local/omni/src/examples/echo/eg2_clt
> corbaname::localhost/NameService#test/eg2

Did you actually try this? Did it work?
I'm actually looking for an easy way to get
IORs for servants registered in the Naming
Service on a remote machine.  But, I don't
have the luxury of NFS, rcp, or scp.
I'll probably, eventually, have to enlist the
help of the Notification Service and possibly
an interface repository.

Regards,

Richard

> 
> You could also just fix the server to bind its object in the proper
> nameservice on start up. Still, nameclt is sort of a neat tool for
> this sort of thing:)
> 
> Your sincerely,
> 
> bjorn
> 
>