Re: Sigtran extension for network management and association changeover.

Dear David,

Thank you for your response.

I want to say there is another primitive, Receive Unacknowledged Message.
If we just retrieve unsent messages, but not unacknowledged messages,
duplication will never happen.
If there is no unsent messages, then the sender will resend nothing.

The duration of association failure detection is another issue, I agree it does impact the changeover performance.

Regards
Xiangsong

----- Original Message ----- 
From: "David Laight" <[email protected]>
To: "Xiangsong Cui" <[email protected]>
Cc: <[email protected]>
Sent: Friday, March 05, 2010 5:52 PM
Subject: RE: [Sigtran] Sigtran extension for network management and association changeover.



> > No you cannot because you risk message duplication which is 3
> > orders of magnitude worse than message loss for SS7.
> 
> I don't understand your comment here, if you retrieve unsent messages
> and resend them, how does the peer get duplicated message?

Because you need to retrieve messages which the far end hasn't received,
not those you have not received an ack for.
In general all messages will have been sent (limited by the TX window),
so there will almost never be unsent messages.

The MTP3 retrieval procedures collect the sequence number of the last
received message through an alternate channel (another signalling link)
so that duplicates are avoided.

Clearly and implementation which tied SCTP and M2PA together could
obtain
the relevant messages from one set of saved buffers - but that would be
hidious layer-breaking.

The main problem I see with using SCTP as a replacement for MTP2 (even
as M3UA - never mind M2PA) is that, at least with the default timers,
the
detection of broken network links is exceptionally lethargic.
SCTP also seems to have a habit of sending data down unverified
(by heartbeat or other response) network links.

David