RE: [wsrm] Proposal for CF4

[Jacques Durand <JDurand@us.fujitsu.com>]

Title: RE: [wsrm] Proposal for CF4

Doug:

agree with most of your improvements below. This section needs a bit of discussion though:

>I suspect the entire list of resend termination cases results in group
>termination for an ordered list.  Is the Receiver RMP required to use the
>GroupAborted fault in that case or should the clarification mention that
>group termination of an ordered group happens in all cases of resend
>termination?

Now that you suggest this, I would favor sending GroupAborted fault (in addition to the precise message fault)
whenever the Receiver considers that further resending will not help to recover from this. This is more explicit...
in all other cases, the groups will terminate by normal expiration of pending out-of-order messages.

Jacques

-----Original Message-----
From: Doug Bunting [mailto:Doug.Bunting@sun.com]
Sent: Friday, July 30, 2004 1:33 AM
To: Jacques Durand
Cc: WSRM (E-mail)
Subject: Re: [wsrm] Proposal for CF4

Jacques,

I must have lost track of all the CF issues and their status.  Back on the
15th I sent an email entitled "Proposed resolutions for ChrisF issues
3,4,6,8,10" that started to cover this and thought we were done.  Your
proposal covers other edits to "line 489" (way back then) which have
happened in the meantime and introduces the need for another clarification.
  Glad you remembered we were not done.

I agree with the direction your update to Section 5.1.3.5 but am unsure
about using a general reference to section 3.2.1 from there, where the
previous update lies.  If we provide the clarification you mention in
section 3.2.1, that reference would cover both the general list of resend
terminations and the specific list of (abnormal) group terminations.  Might
be better to list the specific faults in 5.1.3.5 (provide the clarification
there).  The reference would then go in the other direction -- something
like "In some cases, the group containing this message terminates as well
(see Section 5.1.3.5 for more information)." after the "delivery failure
instead" bullets in 3.2.1.

I suspect the entire list of resend termination cases results in group
termination for an ordered list.  Is the Receiver RMP required to use the
GroupAborted fault in that case or should the clarification mention that
group termination of an ordered group happens in all cases of resend
termination?

thanx,
        doug

On 29-Jul-04 22:07, Jacques Durand wrote:

> Proposal for C.F. comment #4:
>
> agree with C.F. (use MUST), although this is more a matter of
> optimization vs ease of implementation
>
> "A Sending RMP MUST NOT  resend a message for which an RM-Reply with one
> of the following Fault types has been received, and must notify its
> Producer of a delivery failure instead:
>
>             "An Invalid Message Format fault code (Table 22)
>             "A NonSupportedFeature fault code
>             "A PermanentProcessingFailure fault code"
>
> Also, group termination must be updated consequently:
> Section 5.1.3.5 (termination by ordering failure), the Triggering event
> (in both Sender and Receiver)
> should be extended with:
> "...or a [sent] message is faulted with one of the codes mentioned in
> section 3.2.1"
> (replace "sent" with "received", for Receiver)
>
> Another a related issue, when a Sender receives Faults such as
> GroupAborted, or PermanentProcessingFailure, is that not only resending
> but also new sendings for the group should be stopped. That needs be
> made more explicit.
>
>
> Jacques
>