RE: [wsrm] Issue: ACK behavior with ordering and duplicate elimination

> Agree that acks should be sent for any message received, including
> duplicates.
> Note that messages that are resent are duplicates too: yet each of them
> needs be acked
> separately (e.g. the resending may be caused precisely by a lost ack).

Absolutely right. Acks must be reesent even when a duplicate message is
received to ensure at-least-once in case of loss of the first ack.

>
> The semantics of Acks, as described in the spec (1.5 Terminology), is of
> a release of responsibility of the Sender: The Receiver node takes over
> the responsibility of processing the message further.
> (which means, the only thing a Receiver needs to do to ensure this before
> sending an Ack,
> is probably to persist the message.)
>
> Jacques
>
> -----Original Message-----
> From: Patrick Yee [mailto:]
> Sent: Tuesday, August 12, 2003 8:09 PM
> To: Scott Werden
> Cc: 
> Subject: Re: [wsrm] Issue: ACK behavior with ordering and duplicate
> elimination
>
>
>
> >A WSRM receiver must send the ACK after duplicate elimination (in order to
> >not send duplicate ACKs).
> >
> >
>
> IMHO, there is no harm to have duplicate ACKs. In fact, it will be clear
> to the implementations of WSRM receiver that one ACK will be issued for
> each incoming message, regardless of the WSRM features employed.
>
> >A WSRM receiver should send the ACK before ordering (as opposed to when the
> >message is delivered to the app, since this will delay the ACK for all
> >messages queued up waiting for a delayed message in a sequence).
> >
> >
>
> Agreed. Once again, I think ACK could be independent of other WSRM
> functionalities (like ordering and duplicate elimination).
>
> >Comments?
> >
> >
>
> Just my $0.02.
>
> Cheers, -Patrick
>
>
>
> You may leave a Technical Committee at any time by visiting
> http://www.oasis-open.org/apps/org/workgroup/wsrm/members/leave_workgroup.ph
> p
>