Notes from the OASIS
WSRF TC Face-to-Face on
26th & 27th October 2004
Roll call
The roll call is kept on the TC web site
under the meeting record.
See http://www.oasis-open.org/apps/org/workgroup/wsrf/event.php?event_id=5688
Approval of minutes from previous telecon (18th
October)
See: http://www.oasis-open.org/committees/download.php/9879
(IanR) Details of actions on issue 4 and 25 need to be
added.
There were other comments.
Proposed (Ian) seconded (Dave) to approve the minutes
No Objections.
Other Action Review
(Glen) Propose wording to resolve issue 20 (Notification
message format) to mailing list. (Carried fwd from 23rd August) Action:
Drop from action list, but keep on the issue list.
(All/DavidL) Review requirements Doc: (4th
October). David appealed for use cases to support the requirements. Done
(Chairs) Add to the face-to-face agenda a discussion of:
Issue 1: portType aggregation and associated issue 71: Rules for Property
composition. See below.
(Bryan): Move issues 4 and 27 to ‘resolved’. Done
(Igor): Open SetResourcePropertiesDocument issue. Done
(Ian): Write a proposal for issue WSRF 25: More informative
fault messages. (Carry fwd)
(Bryan): Move issue 26 to ‘closed’ Done
(Text editors): Decide which proposal is best to resolve
issue WSRF49 (Pseudo-schema language in the specs) and apply it consistently. See
below.
WS
Resource Spec – Steve Graham
See presentation at: http://www.oasis-open.org/committees/download.php/9878
(SteveG agenda:
è Purpose
è Results
è Points for further clarification
è Implication for other specs
è Dicusssion.
Purpose:
è Clarify “implied Resource pattern”
è Separate concept from embodiment
è Publish a Normative specification
·
Define terms and embodiments.
Result: some definitions
è Resource
è Resource Identifier
è WS-Resource
è WS-resource Reference
SteveG)
(MartinC) Why do we say that the identifier is used by the
WS Resource?
(IanR) It’s because this is less prescriptive about where
the disambiguation happens.
(Umit) A WS Resource is implicitly a reference to the
resource behind, isn’t it?
(SteveG) Well, the WS-Resource Reference is the reference –
and we should be careful about the words.
(TomR) If a box has two interfaces, is the resource the box
or the interfaces?
(SteveG) Like blades, for example. The resource is the
blade.
(TomR) and the resource does not need to know its own
identifier?
(SteveG) No.
(Fred) About equality: if references are equal, they might still
be resolved dynamically into different Web service addresses by different
networks.
(SteveG) Yes, the address needs to be considered in its
context.
(MartinC) If we have a resource which is the blob of state,
and different WSDLs…
(SteveG) The intent of this spec is not to describe types.
(DaveS) Even if I have no resource Propoerties, I must
support the access mechanism?(even if I only want to use WS-transfer)
(SteveG) Yes.
(DaveS) So we are distinguishing this from normal Wsb
services.
(TimB) The ‘properties’ language is not a complete generic
description of stateful services, it’s possible that useful stateful services
will be created that don’t or can’t use ‘propoerties’.
(SteveG) But what would it mean to destroy the resource?
There would be no
(Sam) If we use WS-Resource Lifetime, does that require
WS-RP?
(DavsS) This is an interesting issue for WS-RL.
(SteveG
Results - Embodiments
Points for clarification.
è Resource must/may have properties
(TimB) ‘Must’ restricts the scope of application. Our charter says deal with
generic stateful resources.
Proposed to keep as ‘must’. Only one objection.
è WS-Resource
·
Describe in terms of “WS-endpoint”
·
Must/should on properties and lifetime
è WS-Adressing embodiment
·
Collapse 3.1 and 3.2 in a single embodiment
è WSDL embodiment
·
Clarify 3.4
SteveG)
(IanR) BPEL want to do their own thing with terminating
resource lifetimes.
(DaveS) We need to treat the ‘must/should’ on Lifetime as an
issue.
Action: (see below) Raise must/should as an issue.
(SteveG) Embodiment 3.4 needs further clarification.
(SteveG
The Implications for other specs are:
è Remove Normative refs to state paper
è Prune definitions sections to remove “implied resource pattern”
“WS-Resource” etc.
è Search and replace “implied resource pattern” with “WS-Resource Access
Pattern”
è Appropriate normative attribution to the WS-Resource spec for first
uses of the terms:
·
WS-Resource, Resource, Resource Identifier,
WS-Resource Reference, WS-ResourceAccess Pattern
è Further work on app notes
WSRF Primer - Tim Banks
See presentation at: http://www.oasis-open.org/committees/download.php/9848
The Printer use case based on IPP is a good scenario for the
Primer.
(Igor) It would be good also to illustrate business
application a use cases to widen the scope beyond device management.
(TimB) There is more to the use case than device management,
but yes. Wider scope is better.
Action: (Igor) Construct a ‘Purchase Order’ example.
(TimB) We also need FAQs for the TC Web site.
Action: (TimB) Propose FAQs.
No objections
Metadata for WS-Resources – Tom Maguire
See presentation at: http://www.oasis-open.org/committees/download.php/9822
Proposal document: http://www.oasis-open.org/committees/download.php/9758
(TomM
Motivation:
è Support a broad and evolving range of WS-Resource types
·
Discover descriptions of resources
·
Interfaces
·
Additional Descriptors
è The broader the range of WS-Resources, the more descriptive info is
required.
What is it:
è A Mechanism for additional interface and implementation description
TomM)
(Igor) It would be useful to apply Metadata to a property
description independently of its use in an interface, so that multiple
interfaces and benefit from it.
(TomM) Yes, we should revisit this when we have talked about
derivation of portTypes.
(WilliamV) There are more metadata description tht are
needed – for example describing that a property value sis only available when
the resource is in a particular state.
(TomM) The metadata describes additional contractual
information.
(DaveS) But the WS-resource behaves in exactly the same way
whether or not the document has been read
(DaveS) Shouldn’t xml schema have done all this stuff?
(SteveG) Yes, but we would need to wait a long time to get
into schema.
(DaveS) Does everyone still agree that this ‘separate
document’ approach is the right one?
(TimB) Have we considered how tools cope with two documents?
(TomM) It makes no difference, and there are other
considerations (authorship, cardinality) that are important.
(MartinC) This is all optional stuff, yes?
(TomM) Yes, but we’ll point to it from the RP document.
Tom will continue as editor to move towards a specification.
Others are invited to co-edit.
(See below for formalization of this)
WS-RP on Blackberry Demo – Steve Graham
(WilliamV) How much memory is required?
(SteveG) WSRF takes 19K, Soap takes 40K on top of the
built-in JVM
Issue Resolution
(MartinC) What is the status if the WS Resource document?
(IanR) It’s an editors’ draft. We need to make it consistent,
make it a working draft and post it with the other four specs.
(MartinC) Can we start work on the working draft today? What
shall we do with the ‘WS-Resource’ issues raised this morning?
Proposed (IanR) Seconded(MartinC) That the WS
Resource document be adopted.
There were no objections.
The three issues will be put on the issues list (see below).
The order of business is editorial issues, WS-Resource and
then WSRP.
WSRF 49: Format of Pseudo schema
Proposed: to use the text from the new WS-Resource document
No Objections.
Action: (Editors) Rework drafts by 11th
Nov
WSRF 53 Stop using ‘wsrp’ etc.
Use WSRF-RP instead and make all other namespace prefixes
consistent.
Proposed to resolve to do this. No objections.
Action: (Editors) Rework drafts by 11th
Nov
WSRF56: Specification drafts refer to some non-public urls.
We need to replace the urls with publicly – accessible ones.
Proposed to resolve to do this. No objections
Action: (Editors) Rework drafts by 11th
Nov
WSRF 7: WS-Addressing is not a standard
(IanR) We should drop this issue since we have factored it
out in the WS Resource spec.
No Objections.
Action: (Bryan) Close the issue
WSRF13: Should WS-Addressing elements ‘service’ and ‘port’ be required for
WS-RF?
(IanR) We should say service name and PortType SHOULD (not
MUST) be used, because it might not be necessary for some clients.
(WilliamV) It’s ok not to use the type information if it’s
not needed, but if it’s an error of omission, that’s bad.
A straw poll (who thinks it is important to have an
architectural statement that discovering type information should be possible.)
revealed most people were not desperate for an architectural statement.
Action: (Bryan) Close with no change
We need a new issue:
Examine use of ERPs in service group – in the operational
signatures for add - and base faults specs
Proposed to do as the title suggests. No objections
Action: (Bryan) Move to resolved.
WSRF30: Not clear what the SOAPAction (and wsa:Action)
field should be set to
(Sam) .net dispatches based on the Soap;Action. Is this for
the operation in the most derived portType or the extended portType.
Proposed: (Steve) Seconded (Ian) that in each message
exchange defined within any WSRF specification, there must be a Soap:action URI
defined. Any message which implements the exchange must specify the URI.
For WS Addressing, (therefore specified in the embodiments
document) if a wsa:action appears, it must match the Soap:action URI, even if
the Soap:action is not specified in the message.
No Objections
Action: (Editors) Update specs accordingly. (SteveG)
Update embodiments section. (Bryan) move to ‘resolved.’
WSRF40: Is an EPR the right construct to use to reference a WS-Resource?
This can be closed based on the new WS-Resource document.
Proposed to close with No Action. No Objections.
WSRF41: Is there a way to separate addressing a WS-Resource from the rest
of the WSRF specification?
This can be closed based on the new WS-Resource document
Proposed to close with No Action.
No Objections
WSRF46: Finding WSDL given an EPR
(WilliamV) This is a refinement of WSRF13, and can be closed
with no action.
No Objections.
WSRF47: EPR Discovery
This was a placeholder, but can be closed.
New Issue (PS: WSRF74): Web service vs Web service Endpoint
(Fred) Web service is an informal term, it can mean
something that responds to messages, or a collection of these things, or a
‘type’. When we refer to a WS resource, is it a combination of an Endpoint and
a resource, or a type and a resource.
(DaveS) We can be more precise by saying ‘embodiment’, which
eliminates the ‘type’ idea. We can’t be any precise than that because multiple
endpoints might be the same WS Resource (and resource) or different ones,
depending on the implementation.
(MartinC) The use of ‘endpoint’ isn’t meaningful in the
context of a WSDM addressing scheme.
Action: (Bryan) Close without change.
Meeting recessed 18:30 pdt and resumed 9:00am 27th October.
Requirements Document - Dave Levine
Comments on the document:
è RQT8, 10 overlap: Properties exposed vs gettable
(DaveL) Exposed means there is some visibility, such as a notification, but not
necessarily get.
è RQT3,13 Need for scalability of notification
(DaveL) Most of this is WS-N business.
(SteveG) ) I think there are two use cases. We need a bare bones solution for
WS-RF and leave the rest to WS-N
(MatinC) We shouldn’t say anything in WS-RF; we should depend on composability
with WS-N.
è RQT27 Design time description of what to expect from xsd:any. Good use case
would help.
(DaveL) We need use cases to clarify the requirement: how type information is
discovered, and why
Anti Requirements
è No security
è No transactional behaviour
è Any more?
(DaveL) Comments?
(MartinC) We need requirements and use cases for ServiceGroups.
AppNotes document – Alan Weissberger
(AlanW) The project will be transferred to Katy Warr after
this session - we need feedback on the new outline, and volunteers for the
sections.
This will build on the Primer and be a user’s guide rather
than an introduction.
Introduction:
è
Interdependency of specs with each other. WS-RP,
WS-L etc
è
Exploiters, WSRP etc.
è
Terminology
(SteveG) We need to comment on which applications are
suitable for exploitation of WSRF.
(IanR) So where is the distinction between the AppNotes and
the Primer?
(Glen) In WSDL, we have one document. Here is what it’s all
about, and here’s a use case, and here are the advanced topics.
(SteveG) We should make the Primer a prerequisite to the AppNotes
Motivation:
è
WS Applications
è
Use Cases for the WS-RF specs, Potential
applications for WS-RF specs
(AlanW) Yes. Then we need to take material from the state
papers – volunteers?
(IanR) Yes.
Issue Clarification
è
Behaviour of nillable
è
Etc.
(AlanW) Comments?
(MartinC) What’s interesting are the architectural choices
made be existing applications of WSRF – eg WSDM.
Interop Scenario
(IanR) Do we need an interop scenario/event to test the specs
(Glen) A scenario is useful even though we don’t have
(IanR) We need:
è
Scenarios
è
Published endpoints to test
è
An Event
(DaveS) We should scope ourselves to properties and
lifetime. We don’t have faults ready yet.
(We need a scenario)
(GlenD) We need something that corresponds with the Primer.
(WilliamV) Actually this printing thing is a little
sensitive for HP. I need to check it out with my printer people.
(IanR) What about the scenario description? – If Tim and
William can produce the Printer scenarios it in 2 weeks time, we can then start
a subgroup to work on the Interop scenario. Subgroup volunteers send Emails to
Ian, please.
(IanR) For the schedule, we should aim to get endpoints
available by January, and then think about an interop fest.
Action: (Tim/William) Document printer scenarios for
Primer/Interop
Renewable References – Ian Robinson
See presentation here: http://www.oasis-open.org/committees/download.php/9849
(IanR) It is part of our charter to produce a means to cope
with references becoming stale – eg because an endpoint fails. We need to be
able to renew the reference to regain access to the WS resource.
Using WS Addressing as an embodiment example, we can
identify a resolver service in the EPR, and trhe resolver can supply a client
with a new reference. This TC is chartered to specify the resolver interaction.
(MartinC) I don’t think WSRF TC should be doing this. It is
something that would be done by the implementation under the covers.
(GlenD) This feels close to an addressing issue, perhaps it
should be part of the W3C WS Addressing group.
(IanR) Yes, but that may be a long way off: that group does
not have this in their charter.
(SteveG) We could consider this a domain-specific policy to
solve a quality of service issue concerned with resources.
(MartinC) This is not specific to WSRF. How is interop
affected?
(GlenD) If the means to refresh is vendor-specific, the
client will be unable to reestablish contact.
(DaveS) This is derived from an OGSI requirement for
long-lived handles.
(MartinC) It sounds like naming. There are lots of ways to
do this, why standardize one?
(IanR) This is the proposal, if it doesn’t work, we should
refine it.
(DaveS) What about the other embodiments?
(MartinC) This should be standardized above the embodiment
level.
(IanR) But the solution depends on the embodiment.
(SteveG) But the container approach for addressing is not
really the right one.
(Fred) We need to expand this to include selection of
quality of service. Also, the renewal seems to be owned by the creator of an
EPR, why do we have a separate renewal service?
(IanR) It’s because creation is domain-specific, but renewal
is not.
(GlenD) There are other ways to do this (such as a heartbeat
to provide renewal).
(IanR) Yes, but we need to decide on one.
(GlenD) I think I agree with Martin, we shouldn’t do it.
(IanR) In that case we wouldn’t have replaced OGSI.
(DaveS) This is a critical path for the grid space. It
should be done above the embodiment layer, because switching the addressing
embodiment might be a useful function – something that OGSI could do. I need to
consult with Grid people about it.
Action: (IanR) Put the specification on the TC web
site.
(MartinC) We need to understand the requirements.
Action: (DaveS) I will coordinate the requirements.
Issue Review
New Issue (PS. WSRF75): WS-Resources to use WS-RL be relaxed to a SHOULD
(William) There may be other kinds of destruction semantics
than the one defined by WS-RL.
(Fred) What does ‘lifetime’ mean?
(IanR) It means having the properties about ‘scheduled
termination time’ in the properties document.
(MartinC) But all resources have a lifetime, though it
might not be exposed like this.
(MartinC) We could say that ‘A WS Resource may support
WS-RL’; (and whatever happens in that spec is separate.)
(IanR) Seconded. This replaces the text in bullet 3 of section
2.3.
No objections.
Action: (Bryan) move to ‘resolved’.
New Issue (PS. WSRF76): Clarification of WS-A Embodiment styles
(TomR) For the purposes of referencing the style of
addressing that’s used, can we separate the style of explicit reference
parameters vs including it with the address? We need two sections each
representing one of the styles.
(SteveG) But these two things are the same – the client does
nothing different in the two cases.
(Fred) Can we can put a sentence in to textually identify
the style, so that users of a service can know which is being used.
(DaveS) This does make a difference, because a window on the
client pops up when the reference properties are used.
Action: (SteveG) Make up sentences to describe the
embodiment styles.
New Issue (PS. WSRF77): Clarification of WSDL 1.1 embodiment styles
(SteveG) We need to improve the description because it’s
hard to understand. Could the reference property be in a security context
based on a policy statement? Could it be something the client understands and
manipulates into a header. How can a client know what to do?
(GlenD) The second paragraph is complex, but the client has
Web service infrastructure that can follow the WSDL and policy description to
construct the message.
(SteveG) But the embodiment says the service can do anything
it likes. How can we distinguish regular WSDL from WSDL that satisfies WSRF?
(GlenD) Well, either this is a bad embodiment (too generic),
or the embodiment idea is poorly defined. However, this embodiment works. It is
a WSDL contextual reference.
(SteveG) Does the text describe this?
(GlenD) We should really only have one WSDL embodiment,
which describes that reference data can be included in a message, but not in
the body.
(SteveG) The first embodiment is more specific.
(GlenD) Right, so we can include it within the second.
(IanR) The DAIS group is interested in accessing
Data-oriented resources where the client knows the identifier of the
WS-Resource, and just needs a hint from the WSDL about where in the header to
put it. Is this embodiment described well enough?
(SueM) Yes, this embodiment gets around the ‘opaqueness’
issue, and we could look at it.
(GlenD) I would like to see some simpler text which says we
just use WSDL.
(TimB) What mechanism will the DAIS group use to manage the
resource identifier?
(SueM) It will be some database-access client software.
Action: (SueM) Work with Igor to decide whether this
embodiment can satisfy DAIS usage.
Action: (GlenD) Propose text to improve the
description.
WSR9: API to list all properties.
(IanR) At the July 2004 face-to-face the following proposals
were made.
Proposed
a) A new resource property – allowedRPQNames, ie those which will not throw
‘unknown resource’ fault.
This
properties will return the relevant RP QNames and the schemalocation.
(No
objections)
Proposed
b) A new resource property – presentRPQNnames, which indicates a property of
the relevant name in the instance document.
This
property will return the relevant RP QNames and the schemalocation.
(No
Objections)
Proposed
c) A capability - GetRPDecl (QName) - to return the current RP document
declaration ( minoccurs/maxoccurs/nillable etc) for each QName requested.
(No
Objections)
(SteveG) We don’t need (b) since we have the getRPDocument
resolution. And, we need support of this optional.
(Igor) This should be separate operations, not Resource
Properties so that Resource Propoerties doesn’t exploit its own facilities.
(SteveG) Yes, but the property change notification
announcing new properties is useful.
(Igor) We don’t need (c) because WS-MEX can do this.
(Fred) The definition of which values might be allowed is more
dynamic than metadata can describe.
(William) We don’t need (a) because it can be metadata can
do this, and a notification can be created without hanging it on a property. We
don’t need (b), and (c) is also metadata.
(IanR) But dynamic properties (that are created after the
schema is written) aren’t described in metadata.
(SteveG) I propose we defer this issue. It’s complicated,
and we don’t have a really good reason to do it.
(TimB) Seconded.
No objections.
Action: (Bryan) Create a new list of ‘Deferred’
issues.
Follow-On Planning
We need to set a date for the next rev of specs and get
committee drafts of WS-Resource, WS-RP and WS-RL by the next face-to-face.
(DaveS) We also need base-faults.
Action: Chairs
Face-to-face in January: this is proposed to be at IBM in Raleigh
on Tuesday/Wednesday 2nd/3rd February.
Action: (Chairs) to coordinate with WS-N and WSDM
groups and set the date.
WSRF 1: Interface Aggregation
(MartinC) Why is this important?
(WilliamV) It’s because we need to know what semantics are
associated with a message, and these come from the portType.
(Igor) If we have the portType information only, we still
don’t known which portType provided a particular operation.
() We have the operation collision problem.
(Igor) In WSDM, there are URI’s in properties which can
provide the derived information, so we don’t need the ‘derivedFrom’ attribute.
(William) Actually, we shouldn’t be talking about
aggregation because it’s not any different in WSRF than it is in WSDL.
(Fred) We could simply use a Service with multiple ports (and
portTypes) to aggregate function.
(SteveG) WS-I BP says nothing about the port children of a
Service – this isn’t an interoperable solution. Also, what do we do with the
aggregation of Properties?
(William) The real issue is the incompleteness of the
portType aggregation process. We should deal with that part of the RP spec.
(SteveG) That is a separate issue, we should close issue 1
with no action
Proposed (MartinC), seconded (Igor)
No Objections.
Action: (Bryan) Close the issue.
(Igor) The issue can be closed with advice in the AppNote
that the designer of a portType is responsible for making a sensible choice
about the aggregation of two properties which have the same name.
(SteveG) Seconded.
No objections.
Action: Move to resolved
Metadata proposal
(IanR) Proposed: that the Metadata document be
formally adopted into the TC.
No Objections.
(IanR) We need volunteers to work on editing the Metadata
spec:
Volunteers are: TomM, DaveS, SamM, SteveG, BryanM
Meeting closed 17:30 pdt.
Summary of actions
(Ian) Write a proposal for issue WSRF 25: More informative
fault messages. (Carried fwd from 18th Oct.)
(Igor) Construct a ‘Purchase Order’ example/use case for the
Primer.
(Editors) Rework drafts to incorporate resolutions of
WSR49 (format of pseudo schema), WSRF 53 (Stop using ‘wsrp’ etc), and WSRF56:
(Specification drafts refer to non-public urls.). By 11th Nov
(Bryan) Move issues 7, 13 to ‘closed’ (No action.) Move 30
to ‘resolved’.
(?) Raise new issue: Examine use of ERPs in service group –
in the operational signatures for add - and base faults specs – now issue 73.
(Editors) Update specs according to resolution to WSRF 30:
Define ‘Soap:action’ url and update embodiments description.
(Bryan) Move issues 40, 41, 46 & 47 to ‘closed’. (No
action)
(Tim/William) Document printer scenarios for Primer/Interop
(IanR) Put the RenewableReferences specification on the TC
web site.
(DaveS) Collect requirements for renewable references,
including from Grid community.
(Bryan) Move New issue 75: use of WS-RL is MAY not MUST to
‘resolved’.
(SteveG) Make up sentences to describe the embodiment styles
for Issue 76.
(SueM) Decide whether the WSDL embodiment can satisfy DAIS
usage.
(GlenD) Propose text to improve the description of the WSDL
embodiment. (Issue 77)
(Bryan) Create a new list of ‘Deferred’ issues and move
WSRF9 (API to list properties) onto it.
(Chairs) Confirm if the venue/date for next face-to-face is
really IBM in Raleigh NC, on Tuesday/Wednesday 2nd/3rd
February. (PS The WS-N meeting advised this should be noon Wednesday 3rd until noon Friday 5th Feb).
(Chairs) Decide schedule for next rev of specs
(Bryan) Close WSRF1. No action. Move WSRF71 to ‘resolved’.