Web Services Architecture telcon
29 May 2003

See also: IRC log


Present: Abbie Barbir, Chris_Ferris, Daniel Austin, Dave Orchard, Dave Hollander,David Booth, Frank McCabe, Gerald Edgar, Hao He, Heather Kreger, Igor Sedukhin, Katia Sycara, Mark Jones, Mike Champion, Mike Mahan, Paul Denning, Roger Cutler, Sandeep Kumar, Tom Carroll, Ugo Corda, Zulah Eckert, Yinleng Husband


Chair: Mike Champion

Scribe: Chris Ferris


<Daniel> Roger, here's the email: bparsia@isr.umd.edu

<scribe_cbf> mike: reviews good standing guidelines
... mike: chairs gave themselves ai in rennes to start enforcing good standing requirements

<Roger> I'm getting a fair amount of static. From the listening above one might suspect Frank or Abbie.

<scribe_cbf> mike: trying to steer discussion around specific discussion and proposals around text for the spec
... mike: take a stab at redrafting specific paragraphs
... roger: i liked it too
... roger: wants to know if toronto mtg will really happen
... mike: has had two responses that people cannot go

<dbooth> http://www.w3.org/Consortium/Process-20010719/groups.html#GeneralMeetings

<scribe_cbf> daveo: there are rules if someone formally objects to changing the mtg date/location in 6-8 weeks
... mike: suburb where we are going is where the H.S. with the quarantined people are.,..
... mike: is there a strong sense of the group that we should find an alternate venue?
... roger: I would, and soon
... ???: I would
... daveh: would anyone object to an alternate location
... dbooth: same dates?
... mike: y

<mitrepauld> http://www.city.toronto.on.ca/health/sars/index.htm

<scribe_cbf> mike: ibm interested in hosting in N.C.?
... heather: yes, maybe... need to check if there is a conf room
... chrisf: check on armonk and pallisades too
... daveo: set go/no go date for whether toronto is on or off
... roger: 4 weeks notice tight for me
... mike: understand your concern, what can we do?

<Mark_J> I could attend if the meeting is in Armonk or somewhere else in the NYC metro area

<scribe_cbf> dbooth: proposes we find out from heather if we can host and make a decision then
... daveh: coordinate with wsd?
... dbooth: absolutely!
... sandeep: ?? mentioned there is another meeting that may conflict
... heather: wsdm tc looking at a possible f2f those dates
... ACTION: Heather and Chris to see if IBM can provide hosting for the f2f in N.C. or N.Y. and report back by Wednesday
... daveo: do we have hosts for follow-on f2f mtgs?
... frank: didn't SAP offer for november?
... katia: could host at CMU...
... ACTION: Katia to look into possibility of hosting at CMU and report back by Wednesday
... mike: review AIs
... mike: action items to editors, pretty sure that they have not been fulfilled
... ACTION 2: Katia to look into possibility of hosting at CMU and report back Wednesday
... ACTION -3
... ACTION 3 Katia to look into possibility of hosting at CMU and report back Wednesday

<mitrepauld> http://www.w3.org/2002/03/RRSAgent

<scribe_cbf> :)
... :)

<mmahan> but does it awk?

<scribe_cbf> ACTION 2 = Katia to look into possibility of hosting at CMU and report back Wednesday
... mike: sect 1.6, put rest to rest
... mike: make sure we capture stuff from soap and wsdl, need to do better job than we have
... mike: any other action items that need to be recorded?
... mike: have mgt people met recently, where do we stand?
... frank: proposal is already in there
... mike: how could I have missed that:)
... mike: any other TF reports?
... mike: move on to consolidation of the tentative decisions we made at f2f so we can move forward
... mike: lots of drawing on white boards and sharing of ideas
... mike: lead to lots of email after france
... mike: align wsdl conceptual model, service is collection of endpoints, endpoints have bindings, etc.
... mike: distinction seemed to be important between a service and what wsa draft calls a service that is more oriented to real peice of code
... mike: its a real thing
... daniel: in your note, you say you want to align with wsdl
... daniel: is there a conflict?
... mike: several of us who stayed around after the meeting
... mike: we talked about the implementation side
... daniel: seems like both valid views
... mike: right, want to capture that in the document
... frank: what about one endpoint versus several endpoints
... ugo: thinks the change in wsdl1.2 is a good one
... ugo: wsdl:interface is specific context
... ugo: includes a bunch of operations
... frank: does interface include name of service?
... ugo: no, only collection of operations
... frank: not strong enough to have composition
... katia: right
... ugo: old stack had way to express that by service, where you could aggregate different ports of different portTypes

<mitrepauld> targetResource http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/wsdl12/wsdl12.html#Service_resource_attribute

<scribe_cbf> ugo: targetResource as means of identifying that level of aggregation
... chrisf: wsdl:interface supports inheritance
... frank: to use citibank service to access my bank account, need name of service as well as type and operation
... chrisf: service has endpoints which is a binding of an interface and a wsdl:binding to an address... the service has qname identifier
... mike: @resource is way to identify the thing behind it
... ugo: in some cases, there is nothing behind

<mitrepauld> cf: what do you mean by a composite service?

<scribe_cbf> chrisf: what is a composite service?
... frank: pub/sub service, amazon
... mike: do you mean wsdl 1.2 service?
... frank: there are various ways of accessing, that is supported in wsdl1.2
... daveh: collecting series of capabilities related to same resource
... ugo: the collection of wsdl1.2 services in 1.1 was pretty arbitrary
... katia: yes
... ugo: if we want to use @resource to reproduce that old aggregation, we need to be able to deal with arbitrary aggregation
... mike: to equate one collection of endpoints with another as you could in wsdl 1.1
... ugo: does not map to what you said in your note
... ugo: buyer and seller are two services, want to create aggregation of those
... ugo: ...

<mitrepauld> cf: has different understanding of targetResource, load balancing to same resource accessed from different endpoints

<scribe_cbf> daveo: wants queue management
... daveh: regrets he has to leave (just when it gets interesting:)
... daveo: agrees with chris... wanted a serviceCollection, thought that would be sufficient... but really compelling use cases came up
... daveo: two services cannot share same namespace qname, thats why we need @resource
... frank: no compelling reason to say not to have @resource

<mitrepauld> do: targetResource separate from serviceCollection; need both

<scribe_cbf> frank: not for every service
... frank: can have any arbitrary set of properties associated with a service
... frank: a qname does not identify the printer or the cat
... daveO; not what wsdl group thinks
... frank: there are services that operate on an arbitray number of services
... frank: not clear what underlying resource is being operated on
... frank: have to have identifer for service
... daveo: wsdwg has proposed scheme for identifying abstract components
... chrisf comments on IRC like schema NUNs
... mike: ugo?
... ugo: looks like there are two interpretations of @resource
... ugo: in general case, the two cannot be identical

<mitrepauld> frank: same targetResource may be used in separate WSDL files.

<scribe_cbf> mike: didn't understand that when I wrote this
... mike: does their concept of @resource have any ... is it something that the trunk of a web service needs to know about that the tail does not?
... daveo: pushback that frank gives is interesting
... daveo: what are scenarios that this would be useful for
... daveo: printer example
... daveo: printer is a thing, can talk about printer, could have two services that operate on it
... daveo: much more complicated things like bank account, gets weirder
... daveo: might run into problems
... daveo: elaboration of use cases, needs to be presented to wsd wg
... dbooth: concerned about assertion that @resource identifies the thing behind

<mitrepauld> do: service that operates on a collection of printers; (abstract) resource is the collection

<scribe_cbf> dbooth: printer use case compelling, and it makes sense
... dbooth: but unless the wsdl group specifies that's what that defines
... dbooth: defines a relationship between this and something else

<mitrepauld> cf: confusion is around use of term "resource", need to qualify

<mchampion> Chris: The confusion here is aorund use of "resourece" == needs to be more qualified. It's the logical thing that offers the service, not identity of specific thing being serviced

<mitrepauld> cf: resource is the logical thing that offers the service; not same as bank account accessed through online banking service
... cf: atm same service deployed all over the world; logically the same "service"

<mchampion> chris: ATM offers same "service" to bank, services deployed all over world, but logically it's the same service

<Hao> +1

<mchampion> chris: Resource is not the address but the identifier of what is behind the service

<scribe_cbf> frank: atm/bank account example is a different transport to my bank account
... frank: there is a service around my bankaccount, not always the case that there's one resource behind that that I can tie all the ways of accessing my bank account
... frank: give a service a name and be done with it
... mike: concluding that what I have drafted needs to be reworked
... mike: target resource is a relationship
... mike: likes daveo's suggestion to put together scenarios

<mitrepauld> need way to describe resource relationships. all resources are abstract things. service=uri1, thingBehind=uri2, relationship=uri3 (with semantics)

<scribe_cbf> frank: needs to be something that ties the interface, the sets of endpoints and a deployed instances of that service
... frank: wsdl document is description of the service
... daveo: don't want to get into discussion that the endpoint doesn't exist because there is no wsdl
... frank: need to identify the service
... daveo: wsd totally thinks that this is important

<mitrepauld> do: need uri to uniquely identify the service

<scribe_cbf> hao: i agree that we need uri to identify the service

<mitrepauld> cf: can have separate wsdl files, different services, but same thing behind it

<scribe_cbf> mike: n concepts, not sure if it is 2 3 or more, that all have identity
... roger: thinks that something frank said is wrong
... roger: can you hear me now?
... :)
... roger: need uris to access the service

<mitrepauld> hugo's diagram from f2f: http://www.w3.org/2002/ws/arch/3/05/wsa.png

<scribe_cbf> roger: why should I have to
... daveo: that's why they made it optional
... daveo: confused about whether we are talking about @resource or service
... daveo: service has required @name (NCNAME)
... has optional @resource (anyURI)

<mitrepauld> do: namespace name + name attribute + algorithm to combine into URI

<scribe_cbf> service is identified by targetNamespace:@name

<Daniel> Daniel Austin has to leave a few minutes early - take care until next week, all

<scribe_cbf> (Qname)

<mitrepauld> pauld has to drop off now. bye.

<scribe_cbf> roger: agrees with people who say whether it exists whether there is a wsdl file or not
... roger: endpoint url is what you need
... frank: reason you need a name for the service is that uou can have more than one endoint for a service
... mike: would be nice to have someone own this corner of the document
... frank: we may be pushing back on wsd wg on some of these issues
... frank: not sure we understand the wsd proposal yet
... mike: there are things that wsa cares about that wsd may not
... mike: seeks use cases and suggestions

<Roger> We are certainly pushing back on the term "resource".

<scribe_cbf> daveo: think we may be duplicating efforts here

<mchampion> chris: we need to identify what needs to be identified
... chris: one is clearly wsld:service, but we need to identify other things, like deployed service, and sort out which have which URIs
... chris: we may need to be able to identify SPECIFIC deployed services, having scenarios would help

<scribe_cbf> daveo: thinks that the discussion we have just been having, is much like the discussions wsd has had recently
... dbooth: refering to @resource as refering to the identity of the larger service
... dbooth: perhaps a red herring
... dbooth: is indicating a relationship, may be multiple wsdl services
... dbooth: canonical example is printer
... dbooth: may be abstract printer that is implented by multiple printers
... dbooth: might not be a single thing, migt be another related by a different relationship
... dbooth: misleading as representing identity, but indicates a relationship

<mchampion> chris: sometimes we have a service with one thing behind it, sometimes a service has multiple things behind it
... Sometimes you have different aspects of one thing

<scribe_cbf> roger: can you hear me now? :)
... roger: this confuses me
... roger: something needs an identifier if someones going to use the identifier
... roger: seems to me that there are use cases that bind together the differnt interfaces that bind together the same thingy
... katia: seems to me that there are different levels of the things tha this refers to
... katia: the utility of an identifier for an abstraction also woul dbe related to differentiating this service from another
... katia: think about differentiating for discovery
... katia: comparison at more abstract level
... mike: at abstract view, all these differences matter
... mike: lets develop use cases and discuss what needs to be identified in which scenarios
... frank, katia, chris, others: very productive
... daveo: agrees but wishes we were making progress on the document
... adjourned
... I'm logging. I don't understand 'actions', scribe_cbf. Try /msg RRSAgent help

<Hao> quit

Summary of Action Items

ACTION: Heather and Chris to see if IBM can provide hosting for the f2f in N.C. or N.Y. and report back by Wednesday
ACTION: Katia to look into possibility of hosting at CMU and report back Wednesday

David Booth
$Date: 2003/06/12 14:41:34 $