Minutes W3C WS Choreography WG con call held on 8th April 2003, 1pm PDT.

Agenda:

Dial in information (members only): http://www.w3.org/2002/ws/chor/admin#meetings
1.         Role Call
2.         Confirm scribe
The following is a list of recent scribes (in order): Carol MacDonald, Yaron Goland, Daniel Austin, Jim Hendler,  Peter Furniss, Ed Peters, Greg Ritzinger, Leonard Greski
3.          Additions to the agenda
a.     MC: Harvesting activity
b.     Needs for RosettaNet – covered by item 7.
c.     F2F meeting
1.     WW Grainger will host the next F2F. Dates 25th – 27th June Wed/Thr/Fri
4.        Approve minutes
5.         Action Item Review

1.     ACTION: ALL actions required re-submit use cases with business context

2.     ACTION: HH/YL Check connection of mailing (public-ws-chor-comments) lists to bugzilla – no (bring up as agenda item)

3.     ACTION: Hugo to ask for XML Spy licences - done

4.     ACTION: Yves/Hugo setup an editors ML  - to be checked

5.     ACTION: DC to send details to private list – in progress

6.     ACTION: MM to post glossary document on public list for review - done

7.     ACTION: HH will take a look at Monica’s glossary document – done (track)

8.     ACTION: CM will check with Sun about F2F - open

9.   ACTION DC can check with sonic for F2F - open

6.         Discussion on submitted Use Cases (led by authors)
a.               Please email the chairs to indicate your willingness to lead a use case discussion prior to the call. (David Burdett)
7.         The needs of vertical industry groups (RosettaNet etc) – issue item relating to harvesting
8.         Classification of general requirements 
9.         Reusable choreographies and data formats
AOB.
Summary of Actions

Role Call:

 

Chairs:

 

 Martin Chapman

Oracle

Steve Ross-Talbot

Enigmatec

 

 

W3C Staff Contacts

 

Hugo Haas

 

 

 

 

 

Members:

Mike Brumbelow

Apple

Anthony Fletcher

Choreology Ltd

Mayilraj Krishnan

Cisco Systems Inc

David Burdett

Commerce One

Paul Lipton

Computer Associates

Francis McCabe

Fujitsu Ltd

William Vambenepe

Hewlett-Packard

Yoko Seki

Hitachi, Ltd.

Assaf Arkin

Intalio Inc.

Ravi Byakod

Intalio Inc.

Sanjay Patil

IONA

Richard Bonneau

IONA

Eunju Kim

National Computerization Agency

Abbie Barbir

Nortel Networks

Steve Pruitt

Novell

Greg Ritzinger

Novell

Nickolas Kavantzas

Oracle Corporation

Kevin Liu

SAP AG

Steven White

SeeBeyond Technology Corporation

David Chapell

Sonic Software

Monica Martin

Sun Microsystems, Inc.

Carol McDonald

Sun Microsystems, Inc.

Jon Dart

TIBCO Software

Leonard Greski

W. W. Grainger, Inc.

Daniel Austin

W. W. Grainger, Inc.

Ed Peters

webMethods, Inc.

 

Raw IRC log at: http://www.w3.org/2003/04/08-ws-chor-irc

 

Appointment of scribe:

John Dart, Tibco, kindly volunteered to scribe for the meeting.

Agenda Additions:

SRT: one volunteer to lead a use case, David, based on multi-party case

SRT: stick URL in IRC  http://lists.w3.org/Archives/Public/public-ws-chor/2003Mar/0216.html

SRT: RosettaNet is on the list

MC: Want to talk about harvesting of existing choreography like languages

DA: I wanted to talk about the same things as Martin, in terms of harvesting

DA: I wanted to talk about the F2F

Minutes Approval:

Draft minutes at http://lists.w3.org/Archives/Public/public-ws-chor/2003Apr/att-0093/Minutes_1Apr03.htm

SRT: Any issues with minutes?

Some names are missing from the Role call.

SRT: Can anyone not on attendee list send email?

Mike: was 10 minutes late, not on line

DC: can please use name "Sonic Software" not Progress?

SRT: will circulate minutes again once attendee list is updated

Action items to review:

 

1.     ACTION: ALL actions required re-submit use cases with business context

2.     ACTION: HH/YL Check connection of mailing (public-ws-chor-comments) lists to bugzilla – no (bring up as agenda item)

3.     ACTION: Hugo to ask for XML Spy licences - done

4.     ACTION: Yves/Hugo setup an editors ML  - to be checked

5.     ACTION: DC to send details to private list – in progress

6.     ACTION: MM to post glossary document on publlic list for review - done

7.     ACTION: HH will take a look at Monica’s glossary document – done (track)

8.     ACTION: CM will check with Sun about F2F - open

9.   ACTION DC can check with sonic for F2F - open

 

 

 

SRT: 9 actions from last meeting

SRT: one was re-submit use cases with bus context

SRT: will update mine

SRT: sent out a summary of use cases. Very difficult to find out use case stuff if subject line not prefixed. Please do this.

SRT: Please add "use case" to subject line even if that wasn't the original subject

SRT: checking the condition of the comments list from bugzilla

Hugo: not possible yet it seems

DA: need to bring this up as an agenda item if bugzillza will not work for us

SRT: XML-Spy issue is done

SRT: Yves, and Hugo, you were going to set up editors' list

YV: Will check it out

SRT: DC was going to send details of Stylus XML Editor

DC: Sent it today but it was hung up, should appear momentarily

DC: I sent it to member-ws-chor

SRT: Glossary terms published, thanks to Monica

SRT: Action #7, why was Hugo going to look at glossary? 

MC: Hugo is editor of glossary

LG: this was to look for overlap

Hugo: this is what I found. Defn of web service is different. Defn. of interface seems different, also discrepancy. re service & operation

SRT: We'll have to bring this back as another agenda item

Monica: received a couple of comments from ?? 

Hugo: will send Monica written comments.

Next F2F Meeting

SRT: Re F2F. DA, any report?

DA: Grainger has kindly offered to host F2F in June

DA: Proposed dates are Wed/Thurs/Fri, last week in June - 25, 26, 27

DA: 2 1/2 days of meeting, Wed Thurs + 1/2 of Fri

DA: Will be held at Grainger HQ in Lake Forrest, IL

SRT: Carol & DC, we needed you as backup

SRT: Re dates, can't do earlier because of Java One

??: dates conflict with WS-I plenary

CMcD: you want us to still check out Burlington?

DA: We can move date if we need

SRT: If anyone has issues with date, please let Martin, SRT, & DA know

?? I have to be at WS-I as well

SRT: Might have to be a week later

LG: A week later is 4th of July in the States

?? How about week before?

?? Or Tues-Wed-Thurs of July 4?

CMcD: JavaOne is June 10-13

DA: Please pursue Sun as host - we may need for other meetings

SRT: May need extra meeting this year. Plus there is next year

SRT: Carol & Dave, F2F status?

DC: I didn't get too far

KL: SAP is on for hosting

CMcD: no response from Sun so far

Greg R: I looked into Novell hosting in Waltham, Mass. Could do in June.

Greg R: 1st week in June, 3rd week in June are open. Is booked week of 9th & 23rd-27th

Martin: best is probably 3rd week

Harvesting

 

SRT: Harvesting. MC, tell us about harvesting

MC: Coming from a different angle, one thing we did in ws-arch was a harvesting exercise

looked at ebXML, CORBA, DCOM, etc harvested useful stuff

we can start matching use cases to features, this feature is used in this kind of language

CMcD: Harvesting sounds like a good idea

MC: could put on agenda for another time

MC: could start accumulating lists of things to harvest

?? A good idea

DA: I was going to talk about harvesting. Could talk about harvesting use cases themselves from other groups. Architecture scenarios, WSDL scenarios, from which we could gather use cases

SRT: I agree

SRT: I don't know who looked at summary, but until I saw & went thru it I didn't realize how much we had

DA: from the point of view of being a user & trying to understand, if we can leverage use cases from W3C & use our doc to illuminate choreography. aspects, we could achieve a better level of standardization.

DA: good to have set of shared use cases

DA: maybe should correlate our list with other groups

jdart: may not get sharing of use cases given we have diff charter

SRT: maybe take existing use cases and have good understanding & summary

SRT: will soon find out if we have enough info or not

LenGreski: adding on to MC's point, review of other groups' use cases could help us define boundaries where our use cases shouldn't go.

Carol: how about harvesting use cases to make sure we have those from previous specs

LG: reviewing others use cases could define boundary of where we shouldn't go with our stuff

MC: from my observations of WSA may not go into enough detail

LG: if our work is to be an independent contributions should not have great overlap

DA: we are trying to build an overall set of standards, one group might look at use case from one point of view

LG: maybe we are using the word overlapping differently

MC: need people to drag in these use cases

SRT: there is some harvesting we ought to do per the charter. Some inputs in charter are mandatory inputs. DAML-S, BPML. There are areas we need to look at anyway 

SRT: are 2 buckets. Technical concept harvesting & use case harvesting

SRT: we ought to be able to classify things we harvest into approp. buckets

SRT: we already have a list of things to harvest from

DA: I see 3 diff buckets. Could define someone as subject matter expert, go out & report back to group

DA: could assign so to harvest from use cases, W3C specs, other proposed specs, BPML, DAML-S

SRT: fine idea

MC: need concrete actions: list of places to harvest from: WSA, WSD, Jim's use case stuff

SRT: takes action to issue list

SRT: will send out list

SRT: will take volunteers

jdart: what were 3 buckets

DA: 3 places are use cases, W3C specs, other specs

ACTION: SRT/MC will prepare list of places for harvesting use cases and specs and send list to WG list

Needs of vertical industry groups (RosettaNet etc)

SRT: vertical industry groups. Who suggested this?

SRT: person not on call perhaps

SRT: came out of abstract WSDL discussion

DB: may come out of my use case

[the following occurred later in the meeting but has been here as its is part of this topic. Full chronological order is in the IRC log]

?? Who could be the potential users of the choreography definitions? RosettaNet, etc. are one class of users. Would be good to have outreach to these groups

SRT: could this occur as part of harvesting

MC: Need to encourage user organizations to participate

SRT: add to minutes: identify groups to outreach to as part of harvesting

Tony: I didn't request vertical industry groups to be added to agenda. But what we're really specifying is a language to specify choreographies. Vertical industry groups could be asked for use cases

Tony: when we have a language or at least an embryonic language, could ask user groups if they could use it to define choreography for their particular groups

Discussion on submitted Use Cases (led by authors)

SRT: DB to lead use case

DB: will go thru use case & why is valuable

DB: will look at requirements from use case

DB: use case is a large manufacturer (auto) in us buying from Korean co

DB: buyer has contract with air freight co

DB: typical of just in time delivery in industry

DB: want to minimize stuff to keep costs down

DB: GM wants to control delivery of goods, particularly from smaller suppliers

DB: they can negotiate lower costs by going to a few suppliers

DB: the buyer -- GM or Ford -- placed an order to manufacturer a certain amount of goods based on pre-existing contract

DB: say, we now want you to deliver 5000 units for pickup by air freight co

DB: 3 companies need to co-ordinate pickup and delivery

DB: I describe the sequence

DB: very similar to EDI documents

DB: Korean & air freight co follow same standard as companies in US

DB: This is not restricted to what GM does. If a small co had to do things differently for every customer, may not have IT resources for that. That is one of the problems with EDI

DB: EDI specifies document contents but not choreography. GM says you have to talk to me this way - big suppliers do this because they have to. But smaller guys didn't take part 

jdart: how are smaller suppliers integrated?

DB: manually (fax/phone)

MC: do we need to take account of this in ws-chor? We need to be careful not to confuse web service choreography with more generic B2B issues.

DB: If you can't get the small guys integrated, your ordering times go up, so you need to keep more inventory, so profits go down. So you want to automate as much of supply chain as possible

DB: there are other flows to meet a particular need. Doesn't need to be only one flow, just a small finite number

SP (iona): is the point we want to standardize subset of reusable processes, or process for building processes?

DB: the latter - ability to build processes people need

DB: we can't say to bus guys how to do choreography

MC: agrees

DB: we need to make it cheaper & easier for people to automate, do what they want

MC: this is related to reusable choreography & data formats

MC: what use case highlights is requirement that languages we come up with allow us to define choreography. flexible wrt the things that are being sent

DA: implies is independent of message format

MC: I didn't exactly say that

MC: we need to get to next level of detail to work out exact language

DA: need to have clear MUST & SHOULD

MC: need to turn bus into technical requirements

DA: trying to understand your exact requirement

MC/DA: think we're agreeing

SRT: do we need to represent manual interactions?

SRT: could be just another implementation, just a bit slow

SRT: if you can't tell it's manual and just see message exchange ..

SRT: one of the use cases (from Greg?) was the issue of monitoring

SRT: if we enforce a contract at an endpoint, do we need to monitor endpoint

DA: is it part of choreography to enforce these things, or some other mechanism?

MC: how does one enforce the choreography?

MC: should one enforce a choreography and who would do it

DB: you could imagine talking to someone in Korea using fax & phone. Does this mean you need a variation of the choreography. that excludes interaction w Korea because it uses fax & phone?

MC: we don't want to describe bus process

MC: (versus chor)

DB: this choreography or workflow has been simplified. Could have variations

DB: Between placing order & goods being shipped, buyer changes mind

DB: Could have placed order with multiple suppliers. One can't deliver so increases order with other suppliers

DB: Manufacturer. could be planning to make delivery but there's a breakdown, can't deliver - needs to notify buyer

DB: air freight co may need to change delivery agreement after it's made

DB: nothing's fixed until it's executed. But this is not shown in use case

MC: need state mgmt query on state, change state. Some kind of compensating & consequential actions - recovery

DA: 2 requirements: one is capability for runtime changes to choreography.

DA: one is requirement for conditional logic at runtime

MC: agree

DA: there are two specific requirements in addition to the ones Martin listed

?? was intention to point out that changes were r/t only or can change definition of choreography to use w another buyer

DB: in financial services industry (ATM) - 80% is error checking/recovery

DB: need to handle all failure modes

DA: so requirement for exception handling

MC: exceptions & compensating actions

MC: are we talking about changing participants or choreography on the fly

DB: we are assuming choreographies have dumb, stupid computers behind them and they just follow it. No AI

DA: we managed to route packets, do optimization

DB: one doesn't document the routes, one documents the algorithm. You can't write down potential paths.

 

<music> la la la, tra la lee </music>

  

 

Sanjay: By the way, do we want to capture our understanding of the requirements as we discuss the use cases?

MC: yes Sanjay that would be a good idea, I will harvest from the minutes

DA: Hugo muted me thinking I was the music

DA: when the music came on I was going on a rant about dynamic change in choreography. I do believe there is a requirement for this

DA: this is not just the path you go down when an error occurs, but response to changes in environ. E.g. at 5pm you switch to a new choreography. May at that point store an order till tomorrow. This is very common

??: This is not a dynamically changed choreography, but a higher-level choreography that selects sub-choreographies based on time

DA: I'm quite happy with idea of a master choreography, it solves the problem

??: Good point that it's not just an error condition. But don't want dynamically modified code

DB: does this imply we need a hierarchy of choreographies?

DB: This is a possible requirement

SRT: in the world of composition of choreographies, can you dynamically compose a new choreography and add it to the mix that are currently running

DA: if you have rules to make a deterministic decision

DB: there could be multiple choreographies for placing an order and for paying for it (e.g. credit card, debit, payment terms)

DB: want to be able to compose these

DA: variations on a theme

DB: I agree. Something to think about

DB: 2 last points. Each of 3 businesses - auto, Korea, air freight, wants to manage own IT systems. Are autonomous.

DB: But need to agree on exchange of messages

DB: Last point is that the detailed content of the message can change. If the order is international, need to include customs info. IF it is by air, need air bill describing flight info

DB: But basic sequence of messages does not change

DB: If it comes from Chicago it goes by truck, but the sequence is the same

DB: The documents themselves change. May need line item for customs duties if international, otherwise you don't

 

ACTION: Martin to extract requirements and issues from the minutes

 

Classification of general requirements

 

SRT: who asked for classification of general requirements

DA: I asked for this

close agendum 6

DA: I see 2 kinds of requirements. I am working on outline for requirements. doc. The first ones are use case specific and are derived from use cases. The 2nd class are non-functional requirements --- a terrible name -- dependency on existing standards, technical requirements

DA: Can we classify general requirements?

DA: Scalability, interop, security

DA: There is some overlap to other WGs and other standards bodies. They can't be ignored

MC: Can't we use main categories in WSA as a starting pt?

DA: That was what I was heading towards

DA: I wasn't necessary. thinking about that particular set, but that's a fine place to start

DA: I have this outline

SRT: action to DA to extract classifications. in WSA

DA: will take action

ACTION: Daniel to extract classifications. in WSA

DA: I don't necessarily think can take WSA stuff unmodified

SRT: if you wanted to embellish it, that's fine too

DA: what I plan to do is to publish boilerplate from a standard document, publish that & an outline

DA: how does choreography fit in with other stds

DA: may want to create a dependency map for our group, how we fit in with other stds.

DA: what is the relation between WS-Chor & WSDL, WS-Reliablity, and so on

 

Reusable Choreographies and data formats

 

SRT: I have one item left: reusable choreographies. and data formats

SRT: will need to revisit this next week anyway

MC: this might be one of the cases where we get a technical solution and everyone agrees with it

MC: we have a requirement that the flow is the same but the payload varies. But to have communication at some point you do need to buy into a particular payload format

MC: want to change payload w/o changing flow

?? Are you saying we need to use the XML canonical form

MC: I am thinking more about being able to define choreographies using notion of template types, when you concretize it you fill it what the A's & B's & C's are, or you use an "any" type

SRT: you are talking about patterns of interaction

MC: exactly

DA: the notion of certified or signed choreographies is interesting

DB: I have a preference. One way to do it is to write sequence of messages and define in that message content.

DB: Or can specify a pattern not specify message, extend it and fill in detail (messages)

DB: Or can define flow and then a separate binding document

DA: 2nd one is what we call in C++ a virtual constructor

MC: do people think requirement is a useful thing - abstract messages from flow

SRT: how would name such a thing? The naming becomes an important part of what it is. The pattern of interaction needs to be described ontologically in some sense. So people can find relevant choreographies

SRT: it is a higher-level problem - we may have diff names for same thing

?? Payload can vary based on context

 

SRT: It's 10:29 now. This topic could be a conf call on its own

DB: the whole idea of name resolution - e.g. schema ns to schema resolution -- needs to be done, hasn't been addressed

DA: That's a political issue. Don't want to go into the namespace thing right now

SRT: logging this as an issue

AOB

SRT: Ran out of time, thanks very much everyone

 

Summary of Actions:

 

ACTION: SRT will prepare list of places for harvesting use cases and send list to WG list [1]

recorded in http://www.w3.org/2003/04/08-ws-chor-irc#T20-41-43

ACTION: chairs to compose list of tasks for a call for SME volunteers in next week's call [2]

recorded in http://www.w3.org/2003/04/08-ws-chor-irc#T20-42-02

ACTION: Martin to extract requirements and issues from the minutes [3]

recorded in http://www.w3.org/2003/04/08-ws-chor-irc#T21-18-59

ACTION: Daniel to extract classifications. in WSA [4]

recorded in http://www.w3.org/2003/04/08-ws-chor-irc#T21-22-00