[omniORB] question on usage of omniORB

[Renzo Tomaselli]

Hi all,
    we had similar issues, although not for command line usage. A general
problem class is how to map a symbolic representation of objects vs. one =
of
their "living" implementation, and this becomes very important whenever
administration/configuration issues enter the game.
Such a need appears for large-grained persistent objects, such as service=
s
or other top-level entities.
The solution we adopted has been taken once more from genior/catior tools=
:
we manage <host,port,key,typeID> tuples and then we generate a "syntetic"
IOR through a general building method. This solution offers a much friend=
ly
dataset to maintain in case objects move and it is also much handier for
display/debug/check as compared to blind IOR strings.
A further advantage is the capability to dynamically switch across severa=
l
implementations of the same object according to specific requirements, su=
ch
as in/out-process, a more convenient host, etc.
A minor disadvantage concerns the need to agree on the key, e.g. the real
object "name".
No name service is involved since we never use it.
While not standard we feel this approach cannot be avoided for many real
life applications and I wonder while it never took place in any CORBA spe=
cs.
revision, al least for TCP/IP profiles.
Sincerely,
                                             Renzo Tomaselli
-------------------------------------------------------------------------=
--
TecnoTP s.n.c. Special Information System Design
Maso Pelauchi I38050 Ronchi Valsugana,  Trento TN  ITALY
Tel. +39 0461 773164      Fax. +39 0461 771514
e-mail: renzo.tomaselli@tecnotp.it
-------------------------------------------------------------------------=
--


----- Original Message -----
From: "Donna Maskell" <Donna.Maskell@udlp.com>
To: <omniorb-list@uk.research.att.com>
Sent: luned=EC 4 marzo 2002 22:40
Subject: [omniORB] question on usage of omniORB


> I have seen several postings where people talk about launching their
> applications with command line arguments to indicate the IOR of an
> object reference. This confuses me. I have launched the echo server and
> client on a Unix system, using the IOR output by the server when
> launching the client, and this seems like a very awkward and hardly
> robust interface. On the other hand, I found some code that allows a
> server to bind an object to a name and allows a client to get a
> reference using the same name. Aside from the need to make sure there i=
s
> not name duplication and that everyone who wants a handle to the same
> object uses the same name, this seems a better option.
>
> On the other hand, I don't believe the rest of you are out to lunch on
> this, so what am I not seeing? What is the advantage of using an IOR or
> URI or whatever as an argument? Is it for flexibility? I think my
> applications will generally set up the same set of connections each tim=
e
> they run. I would like to know about the more complicated usage in case
> it does offer functionality I need.
>
> Thank you in advance,
>
> Donna
>
>