On Mon, 12 Mar 2001, Sacha Labourey wrote:
> Hello Jeremy,
>
> I think that Tom's point of view was to allow, for example, to have all AAA
> services to bind under "Myservices/AAA" on each node (and thus having the
> same naming) without having some sort of "Node1/Myservices/AAA"
Yes, it was, but Jeremy's idea has some merits. As he says, it simplifies
deploying to _some_ of the nodes. I'm not sure how that would impact the
way I suggested doing distributed home interfaces, though. Lemme think
about it for a bit...
Tom
>
> > -----Message d'origine-----
> > De : [EMAIL PROTECTED]
> ...
> > bit confusing, it was just a way of implementing this.
> >
> > Can't this public/private distinction be collapsed into a scoped
> > model where
> > each object can specify which nodes of the cluster it wants to be
> > replicated
> > to when it binds? For example, a "private" service would bind to one node
> > whereas a "public" one would bind to all (or possibly a global scope to
> > allow for membership changes).
> >
> > This adds a third possibility where an object could bind to a subset of
> > nodes, say 2 out of 3. This would directly support configurations where an
> > application was only deployed on a subset, for example, without
> > have to poll
> > the node in turn. It would also allow for a master-slave
> > configuration (like
> > Weblogic's SFSB replication) to be implemented on top [to avoid
> > confusion, I
> > emphasize this is a voluntary M/S deployment for a specific
> > service running
> > on top of a federated implementation].
> >
> > Jeremy
> >
> >
> >
>
>
>
--
"If you mess with something for long enough it will break."
- Schmidt's law of engineering
