W3C

WS Architecture F2F (Friday am)
7 Nov 2003

See also: IRC log

Attendees

Present: Mike, Roger, Zulah, Frank, David, Hugo, Jeff, Bijan_(observing), Jacek Kopecky

Regrets:

Chair: MikeC

Scribe: Paul_Downey

Contents


Document Consistency

<mchampion> http://lists.w3.org/Archives/Member/w3c-wsa-editors/2003Oct/att-0031/Web_Services_Architecture_SpecificationA.txt

dbooth: it's meta-comment rather than text for the document

mchampion: we long ago decided not to use the word component

Scribe: discussion about the paragraph "First of all .." regarding cardinality and constraint

Roger: method of constraints should be cardinality or governance

dbooth: has suggested rewording ..

mchampion: do we need to say "what architecture is"

fgm: abstract has some description ..

mchampion: we considered this and don't think it adds anything.

dbooth: an agent realises a service

Roger: service is conceptual

mchampion: a service is an abstraction and may exist without an an implementation
... a service may exist even if we switch implementations

<dbooth> EricN proposed:
... [[
... Can we be clear on the definition a service, that a service only exists if and when if it's
... deployed within an agent?
... ]]

fgm: the word entity with two different meanings

dbooth: for person or organisation

fgm: and for "object" but not an object!

hugo: how about person or organisation : POO

Roger: an agent is a programme

dbooth: how about piece of software

mchampion: entity should only be used for person or organisation

<dbooth> The group didn't understand what EricN meant by:
... [[
... Do we need a standard notation for a (Web services) contract with annotation? So that it
... is presented consistently throughout the document?
... ]]

Scribe: ACTION: mchampion to follow up with eric what does he mean by "standard notation for Web Services contract"
... Action: fgm to check collation order of concepts
... ACTION: fgm to check collation order of concepts

fgm: we don't have a processing model in the architecture - do we need one ?

Roger: isn't specifying the order things happen difficult

fgm: we may want to be more abstract than SOAP, but we don't say "how things work"

dbooth: processing model of what ?
... message exchange is only one part of the picture

mchampion: we discussed a state-event diagram and dismissed it

hugo: the larger picture includes choreography and so addresses processing

mchampion: we don't want to recapitulate SOAP

Eric: the document is inconsistant, i'm trying to scope within what we need to be consistant within. we have a collection of specifications is greater than just SOAP.

Roger: concerned about focus on one given MEP

dbooth: clarification is the order the concepts in the document to based on importance or within processing model

Eric: order should be the as they are introduced in the processing model

dbooth: we had already decided to order alphabetically

mchampion: what is meant by "standard notation of contract" ?

Eric: standard notation system across diagrams

dbooth: consitancy of iconic symbols (eric) of term definitions (roger) across diagrams

Scribe: group brings Eric upto speed on suggestions the group has accpeted

Eric: i have a list of specific inconsistencies which i'll continue to work on ..
... stack diagram and concepts section should be consistent

fgm: distinction between definition and runtime is bogus

<mchampion> Eric: a processing model view ("execution environment") view of the concepts/relationships should somehow align with the definition time view
... ACTION: Chair will schedule time to follow up on the question of whether we should do more run-time / processing model work
... DavidBooth: Eric's "processing model" may be simillar to his "procedure for invoking a Web service"

dbooth: expand description of stack diagram, and processing model (as we have it) ?

mchampion: we're out of time on this.

Service Oriented Model

fgm: our understanding of intermediary is incorrect: it's below our horizon, and role based

dbooth: transport level detail or specific use of the web service

<mchampion> dbooth: SOAP intermediaries are a way of structuring services

Scribe: WRT to section "Message Oriented Model"

fgm: routing is below our horizon

<mchampion> ACTION: Frank will remove intermediary from MOM
... ACTION: Frank will discuss with others how to refactor SOM to incorporate intermediaries properly

fgm: core of SOA focus is on message and agents are outside our scope, but middleware is about structuring a service.

dbooth: not convinced intermediaries need to be in here at all

zulah: intermediaries are important and should be in our architecture

fgm: pen in hand explains intermediaries may be shells of processing

dbooth: if a message is modified in transit, what is the difference between sending a different message ?

<mchampion> Paul: example of HTTP to MQ bridge as one type of intermediary -- lots of other transport-level issues
... Roger: how about "proxy" that does end to end encryption?

Roger: intermediaries are an important part of an architecture

Paul: cites examples of proxies which route and transport and others that add value, e.g. logging and encryption. likes fgm's shell diagram

<mchampion> Paul: Proxies for routing as well; transport-independent correlation is hard!

Roger: questions separation of messaging from service

fgm: our focus is not on routing and low-level concerns

mchampion: transport may provide routing or may be in a higher level ..

Scribe: fgm, Roger: discuss having intermediaries in both models

dbooth: questions how services may be combined, an intermediary be considered as a separate service or part of a pipeline

fgm: intermediaries could be viewed as simple choreography

mchampion: result of discussion is to continue to include intermediaries

RSSAgent: what actions ?

<mchampion> ACTION: Frank will complete action from last F2F to add body/content to mind map and text for MOM

mchampion: anything else on MOM ?

<mchampion> mchampion: suggests we consider the usual diagram showing the architectural differences between SSL encryption between nodes and SOAP-aware end to end encryption

Roger: doesn't like the explanation of "action"

dbooth: deletes the philosophical definition

Roger: likes the word "program"!

hugo: service description is not linked with message description concept

fgm: choregoraphy identifies messages, WSDL describes their content

dbooth: WSDL does describe MEPs
... deletes message description language section

fgm: Message identifier has been commented out ?

hugo: we have an identifier concept, so we don't need a message-identifier concept as well

fgm: what is the cardniality of message and message-identifier ?

hugo: not every message has an identifier, you could argue they all should have one.

Roger: concerned about REST model terms in resource model, in particular the PUT action

Zulah: i have issues between the connection with resource and service models

dbooth: editorial question re term usage of "service requestor" and "service provider"

<mchampion> ACTION: chairs will determine what we decided on refactoring the SOM at the North Carolina F2F, contact the owner of the action item to see if it will be completed quickly, and reassign it if not

dbooth: ambiguity between agent and organisation

<mitrepauld> i'm online now

dbooth: suggests remove term "service provider"

zulah: service provider is a commonly used term and should appear

<mitrepauld> +1 to zulah

dbooth: wants to be precise

<mitrepauld> Roger ... provider entity, provider agent, service provider need to be clarified

Roger: "provider entity" and "provider agent" and use "service provider" in glossary

Scribe: ACTION: dbooth to clarify term "service provider" and "service requestor" and expand glossary
... lunch - sushi!
... ACTION: dbooth to look at security notes put on public list by Roger

<mitrepauld> Will you be resuming after your sushi?

<hugo> ACTION: Hugo to set up a poll about length of f2f in January

<mchampion> paul, there are only a couple of people left. If you have any comments/suggestions, we can talk.
... I have to leave in about 1/2 hour (12:30 PST)
... ACTION: Hugo to talk to Janet about possible promotion of our Note

<mitrepauld> Mike, I'll look through IRC log. I'll stay connected for a while longer, and chime in if I have something to ask.
... URL above is forbidden. Can Hugo open it up?
... bye

Summary of Action Items

ACTION: Chair will schedule time to follow up on the question of whether we should do more run-time / processing model work
ACTION: Frank will complete action from last F2F to add body/content to mind map and text for MOM
ACTION: Frank will discuss with others how to refactor SOM to incorporate intermediaries properly
ACTION: Frank will remove intermediary from MOM
ACTION: Hugo to set up a poll about length of f2f in January
ACTION: Hugo to talk to Janet about possible promotion of our Note
ACTION: chairs will determine what we decided on refactoring the SOM at the North Carolina F2F, contact the owner of the action item to see if it will be completed quickly, and reassign it if not
ACTION: dbooth to clarify term "service provider" and "service requestor" and expand glossary
ACTION: dbooth to look at security notes put on public list by Roger
ACTION: fgm to check collation order of concepts
ACTION: mchampion to follow up with eric what does he mean by "standard notation for Web Services contract"

Minutes formatted by David Booth's perl script: http://dev.w3.org/cvsweb/~checkout~/2002/scribe/
$Date: 2003/11/13 04:10:05 $