RE: more on collaborations, contracts and 2-27 minutes

[Stefano POGLIANI <stefano.pogliani@sun.com>]

David,

 

    thanks for this mail.

I think that the situations that will be modelled are different one the other. In your case, the ordering is (as I have well understood) across the transactions (well, it all depends on how the sitatuion is modelled, after all).

I had in mind a more simple situation where within the same Collaboration, there are many messages; in some way I cannot receive a confirmation of shipment before having received the order itself.

 

Whilst in your situation the application should play a more important role in the verification process (I think that the kind of verification you describe cannot be done by the middleware but is actually "legacy application" dependent since it is the legacy which has the knowledge of the sequencing), in my situation the middleware is the one responsible for verifying the correct sequencing of messages within the same BP instance.

 

Best regards

 

/stefano

-----Original Message-----
From: Welsh, David [mailto:David.Welsh@nordstrom.com]
Sent: 06 March 2001 15:45
To: 'stefano.pogliani@sun.com'; James Bryce Clark
Cc: ebxml-bp@lists.ebxml.org; Karsten.Riemer@east.sun.com; linkage@interaccess.com; sharma@netfish.com; Dale_Moberg@stercomm.com; duane@xmlglobal.com
Subject: RE: more on collaborations, contracts and 2-27 minutes

Stefano,

There's probably a few different use cases for 'sequential counting' , but one of them that is really a sort of 'crude EDI 101' is where one company's applications (in a binary relationship) try's to do a reasonable business control by itself to look for sequential number sequencing from the sender's applications; cause who say's it's a perfect world.

I've regularily seen this type of thing happen often, especially in situations as an example where a warehouse is sending a logistics planning system it inventory database adjustments (increment/decrement on-hand) and to make sure the systems are in as much sync as possible the receiver application itself imposes sequential numbering on the sender application itself.

If those inventory adjust records get inadvertantly posted (just once) in the wrong order at the receivers end then literally "millions of dollars of damage" can occur; and it ain't easy to undo (trust me if you've ever dealt with company auditors).

 

I think you're right there needs to be "a correct sequencing" service going on by a "middleware" as the 'transactions come flying by' but middleware usually can't determine the complete correct logical context of a transaction being sent, before it's posted at the receivers side in it's database, so we sometimes have to help the receiver applications.

 

Dave

 

-----Original Message-----
From: Stefano POGLIANI [mailto:stefano.pogliani@sun.com]
Sent: Tuesday, March 06, 2001 1:16 AM
To: James Bryce Clark; Welsh, David
Cc: ebxml-bp@lists.ebxml.org; Karsten.Riemer@east.sun.com; linkage@interaccess.com; sharma@netfish.com; Dale_Moberg@stercomm.com; duane@xmlglobal.com
Subject: RE: more on collaborations, contracts and 2-27 minutes

Hope not to miss the point completely.... in which case I apologize.

 

The verification of the correct sequencing should be done by the middleware (the infamous BSI) which should be configured out of the content of the CPA (which, in turn, references the Collaboration choreography specified by the BP).

 

/stefano

-----Original Message-----
From: James Bryce Clark [mailto:jbc@lawyer.com]
Sent: 05 March 2001 22:05
To: Welsh, David
Cc: 'James Bryce Clark'; ebxml-bp@lists.ebxml.org; Karsten.Riemer@east.sun.com; linkage@interaccess.com; sharma@netfish.com; Dale_Moberg@stercomm.com; duane@xmlglobal.com
Subject: RE: more on collaborations, contracts and 2-27 minutes

Dave asks for confirmation that something in ebXML can tag and count messages sequentially, so as to enable crossreferencing, detection of blips arriving out of sequence, and the like:

 "I can understand that at a lower level TRP service, there's something dealing with 'message counting' but I'm not aware that something like this info finds it's way up the stack so the business app's can also do their counting; which does sound like redundant but it is the way of the world and as such it's not going to stop. "

In our "simple negotiation pattern" test, we also noted that messages need to be able to cross reference each other, and sought confirmation of some kind of unique message ID.

Good questions.  I don't know the answer.   Can we do that?   Goes on my list for tomorrow's metamodel meeting.   Jamie

James Bryce Clark
Spolin Silverman & Cohen LLP
310 586 2404    jbc@lawyer.com   

James Bryce Clark

Spolin Silverman & Cohen LLP

310 586 2404    jbc@lawyer.com