Arvola Chan wrote: >
> David and Chris: >
> I think Chris' counter proposal is consistent and workable. I have
> previously suggested similar changes (such as requiring all participating
> MSHs to adopt the reliable messaging behavior if the QOS specifies
> RM, sending DR reliably so that they will not get lost, avoiding the
> complexity of having to retry due to DR timeout, not requiring DR
> for the DR etc.) Woo hoo! I really don't think that it is asking too much to require that all nodes adopt RM semantics if OnceAndOnlyOnce is specified. That is, after all is said and done, what the deliverySemantics is supposed to reflect, that the sender wants the message delivered reliably. >
> On the balance, I also think Chris' counter proposal is a little simpler and
> possibly requiring less changes to the 1.0 spec. I also like the fact that
> NRR can be achieved even for an incoming message that has been
> delivered using BestEffort. >
> In order to allow the DR message to be sent reliably but without
> triggering another DR, onceAndOnlyOnce can no longer imply
> that a DR has to be returned. Precisely. >
> It will be great if we can treat next Monday's meeting a voting
> meeting and we can make some important decisions necessary for
> fixing RM and move forward. >
> Regards, > -Arvola > >
