[omniORB] Implementation Repository

Duncan Grisby duncan@grisby.org
Mon Jan 20 11:24:02 2003


On Friday 17 January, baileyk@schneider.com wrote:

> > Yes, the ImR could manage load balancing and so on. To avoid the ImR
> > being a single point of failure, it could be replicated itself, and
> > produce IORs containing endpoint information for all the ImR replicas.
> 
> Indeed, if one could rely on all client orb runtimes to know what to do
> with these multi-endpoint IORs.  Alas, right now I can't.

True.

> Will the shared state of these ImR replicas reside in a database, or a flat
> file?  Designing the persistent data format should be an early step,
> perhaps right after the IDL has been agreed upon?  I haven't looked at the
> IDL that was posted a few days ago yet.

I think the ImR should work with flat files, to avoid dependencies on
external applications, and to make it easy to install. It could have
an option to use a database, but I think it should work without one.

> Anyone else interested in this, how should we start to hammer out the IDL
> for the ImR and management interfaces?

I guess the best idea is for someone (you?) to propose the IDL in an
email, and have interested people comment on it. After one round of
that, it will probably be time to put a version in CVS, and develop
things from there. For CVS, I can create a top-level omniImR tree in
the main omniORB CVS repository, or we could create a whole new
omniImR SourceForge project. Having its own project would allow more
flexibility in managing it separately from the rest of omniORB.

I suspect that these discussions will generate quite a lot of email,
so I would suggest moving the discussions to the omniORB-dev list. I
bet you didn't know there was an omniORB-dev list... I had it created
a while ago for discussions of omniORB 4.1 development (which hasn't
quite started yet...) but we can use it for omniImR too. The
subscription page is here:

  http://www.omniorb-support.com/mailman/listinfo/omniorb-dev

I recommend that anyone interested in helping with omniImR (or omniORB
4.1, once I get my act together) subscribes to the new list.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan@grisby.org     --
   -- http://www.grisby.org --