Re: [ebxml-dev] <BinaryCollaboration> in eBPSS

[bhaugen <linkage@interaccess.com>]

From: Jean-Jacques Dubray
> You are raising an important issue, which I think can only be resolved
> if we either make the notification failure part of the business
> transaction protocol rather than a separate business transaction as it
> is today, or by making an explicit notification of success (either
with
> a separate business transaction, or as a timeout). I personally favor
> the former solution.

I think success can be discerned reliably from
the transaction protocol without a separate notification.
Failure could be more problematic, especially
where the last acknowledgment does not arrive
due to technical difficulties.

In RosettaNet, the precursor to the BPSS
transaction protocol, Notification of Failure
is a separate transaction that is required
in many PIPs (Partner Interface Processes).

My understanding is that RNet went over all
the design alternatives we may consider here,
and selected a separate-but-required transaction
because it may need to go via completely
different delivery mechanisms.

In other words, if the end state of the transaction
is not aligned properly because of a technical
failure on the last acknowledgment, another
notification using the same mechanism
will most likely fail, too.

-Bob Haugen