Re: CVS HEAD version required for -ORBInitRef ?

[Michael Rumpf <michael rumpfonline de>]

Hi,

Fulko Hew wrote:
> 
> Daniel Bruen <dbruen techfak uni-bielefeld de> asked:
> 
> > I forgot to mention that I am using ORBit 0.5.7 in my last mail...
> >
> > In order for the command-line parameter -ORBInitRef to work, do I need
> > to install the cvs head version of ORBit?
> 
> I am curious to find out how this whole thing works too.
> I can't find any good book (online or off) that describes the mechanism.
> 
> According to the FAQ, "-ORBInitRef" is only supported by the cvs head
> version, but the head version is no longer a work in progress and the
> ORBit2 version superceeds the head version.
> 
> In addition, (and here is where you might be able to help) is, in either case
> how to I get access to the CVS, where is it, and what do I have to
> set my CVSROOT to, to get it, and whats the password for readonly access?
> (I can't find this info anywhere.)

The link on the ORBit FAQ on http://orbit-resource.sourceforge.net is
"anonymous CVS access" (question 2c):
http://developer.gnome.org/tools/cvs.html
 
> Also, I am curious to find out how this whole thing works too.
> I can't find any good book (online or off) that describes the Namming mechanism.
> and how the IORs are exchanges across boxes so that remote machines
> can find them.

The corbaloc/corbaname mechanism is quite new, so that recently
published books do not cover the topic at all. The best article I found
is from one of the CORBA guru's Michi Henning. The article was written
in response to CORBA beginners asking the same question over and over
again: http://www.ooc.com.au/staff/michi/binding.txt
Because it is quite old, it does not cover the corbaloc stuff. The
article describes only the traditional approach, i.e. having IORs which
need to be exchanged by server and clients.
Because it is somehow inconvenient to exchange such an IOR the chosen
solution (by the OMG) is to make an initial reference human readable.
The corbaloc allows a developer/administrator to create such an initial
reference without having to get his hands on a running server instance
first.
He just writes "corbaloc::myserver/ObjectKey" at the side of the client
application. The client uses the information, provided by the URL, to
send a "LocateRequest" (defined by IIOP) message to the server (OK, this
is what the spec. is asking for here, but there are certainly other ways
to implement it....;-). If you don't specify the port the client ORB,
which is responsible for sending the message, uses a default port
number. The response ("LocateReply") from the server ORB contains all
the information which is required to get a full CORBA object reference
for the specified server object. The ORB handles all the stuff
transparently. You just write
string_to_object("corbaloc::myserver/ObjectKey") and you will get a
normal CORBA object reference. What is going on behind the scenes
shouldn't be of any interest, except if you are trying to write an ORB
your own.

BTW, Michi Henning's article is also linked from the ORBit FAQ: Question
1a.

Hope this helps,

	Michael