Minutes of W3C XML Protocol Working Group teleconference held on 28 March 2001.

1. Roll call

PRESENT 42/36 AUTOMATICALLY EXCUSED REGRETS ABSENT WITHOUT EXPLANATION
Minutes:
Eric Jenkins
Action items:
Henrik Nielsen

2. Agenda review, call for AOB

David Fallside: The abstract model group will present an overview of the
model and an explanation of any outstanding issues.  WG members will be asking
questions. The AM Group will provide a draft at end of the week.  
The XMLP WG will have one or two weeks to comment.
Then we will send a draft to W3C for publication on, or around, Friday the
13th.

Noah:  Question about Travel logistics (deferred to the end of the
telecon).

??:  Are we going to have a discussion about Web services conference?
David Fallside:  Take the discussion to the member list.
    

3. Approval of minutes from 14 March and 21 March telecons

No corrections, additions, or modifications.  Minutes stand as is and will
be published internally and on the public site.
    

4. Review Action Items

Follow up on i41:
Noah: Whether we specify an explicit actor value. Henrik and I received
      input from other members of the group and put together a proposal. 
      Larger issues about what you do and don't know about abstractions like 
      path and endpoint. Propose we defer resolution until after do general
      architecture about meaning of path.

David F: Action is to make sure that we resolve this issue. Defn.s of
         path-target-actor, are these covered in other issue(s)?

Noah: Are all design questions listed as issues?

Henrik: Is it reasonable to say that we plan on getting to it.

David Fallside: If we add the comments that Noah made to i41 and leave it
                as open, then we can deal with those issues.

Noah: message patterns and paths is the dog, and 41 is the tail, concern
      that we address the dog rather than the tail.

New action item to Noah to look at issues list and determine where the dog
is.

David Fallside: If not in the issues list, then we do not have a
                guaranteed way of having it tracked.

Action Item regarding i41 is accomplished:

Action Item for Stuart was accomplished

Action Item for Henrik to write proposal for merging Glossary w/ SOAP 1.1.
terminology wasdone.

Action Item for Jeff Kay. He sent out an email this morning with a further
suggestion about how we should manage the issues list, in particular we should
be specific about pointing to and recording the solution to issues on the list.
Need to ask Jeff Kay whether the current version of the issues list satisfies 
his request. Action Item to David Fallside to query Jeff Kay about this.
    

5. Abstract Model, overview and pending issues

Stuart Williams: In terms of process, we are in good shape to have finalized
                 proposal to put to the working group on Friday.  
                 Mark Jones will take on editor role while Stuart is on 
                 vacation. The mapping to Soap  and the comparisons to ??? are
                 not on the critical path.
                  
                 John Ibbotson has been making good progress on scenarios.
                 Henrik has been making good progress on glossary.

                 Issues list will be revised, many of the issues have been
                 resolved.
                 Change log will be moved from front to back of document.
                 Issue 5 will generate a new minimalist picture to add to the 
                 document, similar to fig 2.1, but without intermediaries.
                 Subgroup has no issues against section 1 of the document.
                 Section 2 (overview of abstract model, picture plus narrative)
                 Minor editorial glitches to be fixed.
                 Major issue to be fixed are the concepts of path and 
                 targeting, plus concept of module and handler.  These remain
                 as topics of discussion, and we plan to capture output of
                 debate in document at a later time. WG may need to form 
                 subgroup to address this.

                 Section3 generated the most controversial discussion.

                 I posted something late last week which made an alternate
                 proposal over request-response.

                 I believe the issue is that if let it go, we would lose
                 causality. I  proposed that it would be possible to 
                 reintroduce notion of causality into one way message. Believe
                 is possible and have posted that proposal.

                 Another change in the document, collapsing UnitData operation
                 into four events: send, receive, status and forwarding.  
                 Forwarding is not creating a new message, since it is
                 conceptually the same message.

                 All of this will make it a slightly shorter document

Ray Denenberg: Is XMLP_Data is no longer part of the model?

Stuart: Correct.

RayD: Are you confortable that there is consensus on this?

Stuart: I've had no feedback that indicates dissent.

Henrik: I would say that the current model is more flexible than previous
        model.  This more unified model supports the concept of dialogs, 
        for example.

David Fallside: Consensus so far has been in the AM subgroup, now this
                needs to go to the larger group.

Stuart: The line of thought that convinced me to make this change was that
        we wanted causality rather than XMLP_Data. My proposal is not simply 
        one way messaging but one way messaging with correlation built in

David Clay:  Does this make it easier to support bindings?

Stuart: Mark Hadley has been working on the bindings section, he probably
        needs to think about this.

Marwan: (Summarized Stuart).  So, the notion of causality would be lost if
        we went to a one way model.  I disagree with this and want to follow 
        up on the mailing list

Noah: I just think this is an example where the abstract work is taking us
      into the core work.  I'm not saying this is bad, I'm just pointing out
      the timing.  I feel it is important to tie into a concrete syntax. 
      There should be limits to degree that we want the abstract group to 
      drive the specification process.

Stuart: Some folks feel that the definition of path needs to be nailed
        down and some folks feel that it is fine as is.

Frank DeRose: With regard to the correlation message ref member, it's not
              clear to me that is reflected anywhere within the xml message

Stuart: not necessarily, will depend on the underlying protocols to do
        correlation

Frank: Is this issue brought out explicitly in the document?

Stuart: I will take as an Action to go look in the doc and add if
        necessary

Section 4 is largely Mark Jones' work

Major issue:  question of path and targeting and the apparent magic between
targeted blocks and how intermediaries

Things that we intend to do before submitting doc to WG:

     -Maintenance of headers and bodies and speak about blocks and remove
     headers and bodies distinction.

     -Switch order of 4.1 and 4.2

     -Intend to highlite, but not delete, the parts of text that have to
     do with soap.

Mark: there may be some small changes in the terminology before the
      document goes out

Section 5:

Focuses on protocol bindings:

Work done by Mark Hadley presents a binding model for the transport side of
the protocol.

This may be another area where the WG wishes to commission some activity.
There has been some discussion about the service module bindings, remove http
specifics, and minor changes about how to nest bindings.

David Fallside:  what is status of section 5.3?

Stuart: Henrik and I agree that the matter of carrying arbitrary content
        is a binding issue, hope to be able to exploit binding capabilities.

(Henrik explained his example about the nested bindings piece).

Mark:  What is the process for AM output?

David Fallside: Modifications will be put into draft by Friday and as of
                Friday WG should consider this document a stable document.  
                WG should review that document (dated March 30) and we will
                discuss that document and we may revise it on the basis of our
                discussions.  We could take a couple diff approaches. A formal
                approach is to create an issues list, although it is not clear
                we need to do that at this point.  WG need to get consensus
                about contents of a working draft.  WG can decide about 
                including AM doc in the spec document in our final deliverable,
                and whether the AM is normative or not.  At the present time,
                the proposal is that this doc will exist as a working draft in
                its own right, but could gointo the front piece of the 
                Spec. document.

Paul Cotton: I would like to consider the possibility that if it is published
             as a separate document and we can't agree how they are linked 
             then we should record issues in the abstracts of both documents
             explaining how they are related to each other and that the WG is
             working on issues about the linkage.

David Fallside: Noah's caveat starts to address that issue though this does
                not go as far as saying that there are formal areas of linkage
                that we attempt to resolve at a further date.

PaulC: Mapping AM document to Soap is one place where the linkage is
       indicated.

David Fallside:  When on Friday will the document be ready?

Stuart:  Need to talk to Yves but should be first thing in the morning, PST.
    

6. Report on template for XML Protocol Module

David Clay:  Report on template for xml protocol module.

Paul Denning and Glen Daniels also responsible for this work.

The tack we took was to ask:

  -Who would define module template?

  -Who would use it?

  -How to combine modules into application?

Need to reconcile this with abstract model, because of many changes,
especially in section 4 of AM.  They have different ideas about, for instance,
who orders the modules.  Need to be reconciled.

Mark Jones: We need to reconcile boundaries of terms in the AM work, the
            Template work and Henriks soap glossary mapping.

David Clay, Mark and Henriks will take offline and present at next telecon.

Paul Denning: Since we decided that RPC is a module, should we define as a 
              separate document?

Marwan:  I think this is a good idea.

David Fallside:  Did you touch on RPC's in your template?

Mark:  No our work is more general but we could test it on RPC.

Noah: I think this is a good direction to experiment with, but we have
      technical and political factors to consider:  Don't want to introduce
      gratuitous differences with SOAP, etc. Let's have the group revisit it.

DavidC:  It might be better to start out with an easier case than RPC.

Marwan: What would be simpler, wouldn't anything else be out of scope?
        Would the proposed digital signature proposal be a module?

DavidC: Maybe, yes.

Noah:  (long monolog) Let's try RPC.

DavidC: I didn't think from the definition of module that we would be
        distinguishing between parts like a body or header. Many issues 
        discussed concerning what we mean by module.

David Fallside: The group should discuss these issues and attempt to model
                RPC with the template and get back to the WG at next weeks 
                telecon.
    

7. Issue Reports

(A) Ray Whitmer:  Whether RPC works as a module, subgroup talked about it
                  and concluded you would find RPC in headers and other part
                  of the document.  RPC, when it exists, tends to shield other
                  services.

The group concluded:

RPC works well as a module.

No concrete proposals about how modules may be permitted to extend error
messages.

No clear deliverables for the group.

Recommend that RPC should be tried out as a module.

David Fallside: I think that the RPC recommendation answer was the main
                deliverable for your group. Are your discussions summarized in 
                the message list?

Ray:  Yes, it seems like a reasonable summary.

David Fallside: I think that the RPC task force has accomplished its
                mission, and the best way to go forward is to test it out as
                a module.

(B)Issue 47 & 48. David Ezell: There has not been much discussion on these
                               two issues. Essentially the soap encoding rules,
                               sitting on top of an RPC mechanism of some sort,
                               the actual wording was called out by Vidur.
                               We suggest that we back off from redefining the
                               encoding rules and point to soap note, and
                               address in a later WG.

Marwan:  Why are encoding rules needed in the first place?

David Ezell: People are using these already and for certain cases have
             high utility.

Paul Cotton: I think there are lots of W3 groups that want these in a
             normative location and that xml protocol is the appropriate 
             location.

Noah: We need to do it to promote interoperability.

David Fallside: We will have to revisit this issue since we are about out
                of time.
    

8. Any Other Business

Novell has offered to host the following F2F in San Jose.

PaulC:  Are you aware that there may be a w3c plenary at that time?

David Fallside:  Send me mail.

David Fallside: We have to reschedule the telecons starting next week, the
                day will still be Wednesdays but the time may change must be
                considered in the next day or so

Noah:  How do we make TGV reservations?

Hervé: There is a web site, but they will not be taking reservations
       until the beginning of next month. From Charles De Gaulle, trains leave
       only a few times a day but from Paris they leave more frequently. 
       Will take this to email.