- From: David Orchard <dorchard@bea.com>
- Date: Tue, 2 Apr 2002 13:37:37 -0800
- To: "'Cutler, Roger \(RogerCutler\)'" <RogerCutler@chevrontexaco.com>, <www-ws-arch@w3.org>
- Message-ID: <011301c1daa0$d2db1440$0100007f@beasys.com>
EDI and Use CasesRoger, 1. I agree. Sequencing of reliable messages should be listed as a requirement, and I for one intend to make sure this is in our requirements. Additionally, we believe that this should be in scope for one of our first working groups. Hence why we supported a routing and reliability WG formation back in November. 2. Expanding on your notion a little bit, one possible set of requirements is that the message sender and receipient containers/message handlers/controls should be a first class object, that messages in the container should have expressible states, and that containers should be able to respond to queries against the status of messages in the container. While I agree with this notion (and other similar notions) in the future, this seems beyond what we can do in a short time frame. Again, this is a clearly obvious step when dealing with asynchronous messaging (these things don't make much sense in synch/rpc), and there are proprietary implementations as well as emerging specifications on this. It's just that we(BEA) think we(web services arch) should figure out how to do interoperable asynch/reliable messaging before we figure out how to do all the control logic around it. Plus it layers nicely as there is fairly loose coupling between the runtime message formats and the container control logic. I've specifically answered each of your points with additional information on prioritization and timing, to continue the process of thinking about ordering and scope of the new working groups. Cheers, dave -----Original Message----- From: www-ws-arch-request@w3.org [mailto:www-ws-arch-request@w3.org]On Behalf Of Cutler, Roger (RogerCutler) Sent: Tuesday, April 02, 2002 11:46 AM To: www-ws-arch@w3.org Subject: EDI and Use Cases I have reviewed the notes I have taken talking to our EDI people (and talked some more) and it looks to me like the UC's presently in place cover many of their concerns, particularly those regarding reliable delivery and acknowledgement, asynchronous communications and intermediaries. Although they are not covered in the use cases yet, I assume that issues involving security concerns that are important to EDI applications (authentication, authorization, confidentiality, non-repudiation, etc) will have use cases. I am, however, a bit concerned about two other issues that appear to be important to EDI. The first may be trivial, the second may be impossible -- but they are both things that seem to be heavy hitters in terms of what is expected from a VAN so let me get them on the table for comment: 1 - Unique message ID and sequencing. It is very important to be able to identify a message uniquely, and this identification is generally contained in the enveloping mechanism. (There are usually other ID's in the body, but this is clearly beyond the scope of the infrastructure). The unique identification is commonly done by the combination of "To", "From" and a sequential "control number". This facilitates queries like, "Did you get message N sent from A to B?" "What messages of those sent from A to B are between N and M?" and so on. In addition, the datetime of message envelope creation (not necessarily the datetime sent) is also important. As in "What messages were created on Tuesday by A?" As I understand it, the sequencing is logical, not a guarantee of order of delivery. However, in some cases (as I understand it, this is unusual) the sequencing may be required. That is, control number 21 is not a valid message unless (or until?) 20 has been received. I think that this may be trivial, but my understanding is that it is very important, so I think it's worth making explicit in case there is a joker in the deck somewhere. Now to the one that may be difficult, impossible or need re-stating somehow: 2 - Tracking. Being able to answer the question, "Where is message N sent from A to B?" and get back an answer like, "It's in B's 'mailbox' but they haven't opened it" or "B opened it on Tuesday" or, I suppose, "It is in transit to B and at the moment it is stuck trying to get through the gateway at C". Now I understand that the Web doesn't quite work exactly that way. It differs from a private, physical line in that one does not know up front how a message will be routed or even (perhaps) where it has been?? Or what happened to it if it disappears (which I understand is uncommon)?? However, it seems to me that there are some things that the infrastructure might be able to deliver in terms of tracking -- like a guarantee that the message arrived or that you will know that it didn't (part of reliable messaging). Or what about tracking what intermediaries a message went through or asking where it is in that process? Sorry, I know that this is a little vague. I think I need some help here figuring out what is possible, reasonable and valuable.
Received on Tuesday, 2 April 2002 18:50:03 UTC