Re: [wsrf] Groups - WSRF TC [28April04] notes[3].doc uploaded

[Tim Banks <tim_banks@uk.ibm.com>]

John Kemp wrote:

>Thanks for this. Is there any chance that we could have the minutes
>appear in plain text in the body of the email, in addition to being a
>downloadable document? It would be nice to have the raw text in email to
>facilitate reading and comments from the list.

I guess it's hard to please everyone with one format, but the master copy
of the minutes has to choose one format and the email announcing the
minutes is automatically generated by the Web site services, and it doesn't
include the document inline. So, as far as I know, the answer is 'no', but
we can simply export the minutes to plain text and post them to the mailing
list. See below :-)

Opinions on the format of the master copy are welcome.

Regards, Tim Banks
IBM TP Architecture & Technology. Hursley, UK.
Phone: External +44 1962 815639, Internal 245639

-------- Reformatted from
http://www.oasis-open.org/apps/org/workgroup/wsrf/document.php?document_id=6619

Notes from the inaugural OASIS WSRF TC meeting
 New Orleans, April 28th 2004


Dave Snelling acted as Convener.

Tim Banks took notes.
Roll call

See web site:
http://www.oasis-open.org/apps/org/workgroup/wsrf/event.php?event_id=4786

Election of committee Co-chairs
Dave Snelling and Ian Robinson proposed as co-chairs. No others were
proposed, and
there were no objections.
Welcome and Overview of OASIS - Jamie Clark (OASIS)
Dave and Ian received copies of "Robert's Rules", but most work of the
committee is
expected to operate intuitive and consensually, without the need to resort
to the OASIS
rules.
Review  of Charter - Ian Robinson
Q (Connor): What is the nature of the Royalty Free (RF) status? - There is
more than one
kind of 'RF', so the charter statement should be clearer.
A (Ian): We have a statement of intent about RF, but will treat claims over
IP from
contributors as they arise.

Q (Anish): Is an 'RF' statement about contributions obligatory?
A (Jamie): Declaration of no claim is implied by not stating a claim.

Charter description of the Scope of the group.

Q (Glen): Some of the scope bullets seem to be about implementation.
A: No, that's not the intention, but we need to refine the charter if it is
ambiguous (After
a lot of effort already).

Q: There should be a definition of what a Stateful Resource is?
A (Ian): Yes, there is a need for such a definition - accepted as an Issue.

Q (Savas): What are the constraints on operations with other TCs and
standards?
A (Dave): This is too big an issue to be dealt with in the charter, but we
do need to define
this.

Q (Glen): Are we allowed to clarify the charter?
A (Ian): we can constrain it, but membership of the group denotes
acceptance of the
existing charter.

Q (Richard?): Is the scope Information Ontology agnostic?
A: The stateful resource must be representable in xml, apart from that,
there is no
constraint on the information model.

Q (?): Is the semantics of transactional update in scope?
A: No. transactions are part of QoS and bindings, which are not part of the
business of
the TC, which will deal with the meaning of the single operations.

Q (Phil): Why is the factory pattern out of scope?
A: The prior work (OGSI) discovered too many domain-specific patterns with
better
semantic value.

Charter description of deliverables.

Q (?):  Where did the requirement for renewable references come from?
A: From the OGSi work.

Q ( ): Can the details of these requirements be included?
A: This is too much detail for the Charter

Q (Savas): Will there documentation for the requirements?
A: There is nothing chartered, but there could be a proposal.

Q (?): Is security in the scope of the TC?
A: Not, but it needs to be composable with what the TC produces.


TC  Process
*     Teleconference times: 1 hour, bi-weekly, Monday, noon in East Coast
time
*     Face To Face meetings: July 26 UK, October 1, US
*     Decision Making: Issue-list driven, decided by consensus, backed by
OASIS
ballot if necessary.
*     No Subcommittees

Contributed Work
Resources - Steve Graham
Motivation:  Resources exist - that what the 'R' stands for in 'URL', but
the approach to
it is application-specific. A standard approach helps for
application/domain integration.

The scope of WSRF is identification, modelling, lifetime of state.  Web
services are
typically stateless.

Q (Anish): What does 'stateless' mean. Web services are not supposed to say
anything
about implementation.
A: Sometimes implementation shows through. We need to acknowledge and
describe it.

WS-Addressing is a proposal as the way of identifying resources.
Q: (Connor) Is WS-Addressing in a standards committee?
A: This question is on the issues list.

A WS-Resource is a pairing of a Web Service and the resource.  It has
identity, lifetime,
state, and Type.

Cardinality: a resource can be accessed by multiple services. A service can
access
multiple resources.

Q (Anish): Is there a way for a client to know whether two services access
the same
resource?
A: No - this is an application specific issue.
A(William): the WSDM want to define sameness for their application domain.

ResourceFramework Overview - Wiliam Vambenepe
Besides the identification, modelling etc issues, the specs also  define an
interaction
patterns

WS-Resource Properties - Steve Graham
WS-ResourceProperties defines a scheme to model elements of resource state.
And the
rationale is a basis for standard resource inspection, monitoring and state
management.

ResourceProperties can be used to model attributes of (eg) a Process or
Operating System,
in the domain of systems management.

The atom of manipulation is the xml element which is the direct child of
the root.

Q (Fred): The properties document is not materialised through the
interface. The
structure of the resource is not visible.
A: Yes. Some aspects of the schema (ordering) are not defined, but others
(cardinality)
are.

Q (Fred): Is it true that some schema constructs are not possible in
WS-Resource
properties
A: This is an issue on the issue list. Choice is not possible. Sequence is.

Q (Sue): What about operations which make data inconsistent if there are
two schemas
for the same underlying state.
A: Yes. These are issues for implementation.

Q (Glen): Semantics are poorly defined for Web Services in general. We
shouldn't do
this in the TC, but we need to define the semantics of the messages
relevant to properties.
A: It is an interoperability issue to define semantics consistently, and
compose with other
semantics (transactions) consistently.

Q (Richard): Is there any definition of the read-only accessibility of
certain properties
A: There is a need for metadata (policies) on the properties to describe
their visibility, but
these specs don't deal with it.

Properties are an attribute associated with the WSDL portType.

Q (Anish): Should we be exploiting WSDL 2.0?
A: Yes, we could make it exploit WSDL 2.0, but only have a spec and tools
for WSDL
1.0, so this is the first priority.

Q: (Fred) Is the resource properties value multi-valued
A: No.

Q: (Richard) Is there way of qualifying (subsetting) properties to
represent different
devices.
A: Another issue for metadata.

ResourceProperty Operations are get, getMultiple (variously useful for
different
optimisations), Insert, Update, Delete for elements. The more comprehensive
way might
be xPath: the proposals are simpler, to avoid colliding with xPath.

Q: Is there a need for expressing a relationship between resources.
A: Interesting, but it isn't part of this TC.

Q (Richard): It's possible for a resource to reference other resources,
isn't it?
A: Yes.

WS-notification can be used to set up subscriptions to receive messages
about changes to
the resource property values.

Resource Lifetime - William Vambenepe

Provides standard ways of destroying resources.

Q (Anish): Couldn't we have a canonical way to create resources.
A (SteveG):  it can be raised as an issue, but a use case is needed, since
OGSI did not
discover one.
A (Ian): There has been a long debate over this issue. The minutes would be
useful to
answer this (frequently asked) question.

Q (Richard): This seems rather (overly) complex.
A: Yes, but are/were use-cases driven by shared resource allocation.

Q (Fred): Why is this an explicit op, rather than a get/set of a property
element.
A: The operation has more semantics, such as faults, and termination time
in the return
message.

Notification of destruction.

Q (Dave Martin): Is there any need for clock synchronisation in the
specification of times?
A: No. That requirement has been removed.


Issues with Referencing and Resource Identification -  Anish
Karmarkar

References identify Web Services as targets and as message parameters.
Proposal: "An
EPR cannot be used to reference a Web Service".  The EPR reinvents
something like the
WSDL service element, but is insufficient. It contains a URI which is the
end-point for
messages, but this is not the whole story - it doesn't reference the WSDL
document.

Should references have tightly-bound types? We could use WSRef instead &
create a
new SOAP header block which can be dealt with by infrastructure. The WS-
MessageDelivery spec (proposed to W3C) describes how this would work.

Q (Steve Tuecke): Aren't we redefining what WS-Addressing does: tells how
to take a
message parameter (EPR with end point and resource descriptions) and turn
it into a
SOAP header?

Q (Glen): Isn't this an issue for the wider Web Services Community. We can
deal with
other issues in the TC.

A further proposal: identify the resource explicitly in the WSDL. We need
to be more
WSDL-centric, and not invent other things.


Base-Faults - Steve Tuecke

Motivation: structure for understanding faults would enable construction of
tools, loggin
of events, problem determination etc The standard components are Time of
thefault,
originating service, error code, description.  Faults are stackable.

Q (Glen): can't these issues be dealt with in WSDL when WSDL 2 is finished?
We'll be
left with two ways of doing it.
A: Yes, This is an issue which should be examined in the light of WSDL 2.

Q (Chris): SOAP 1.2 has fields for faults. Why can't these be used?
A: This lacks the facilities of WSDL eg hierarchical definitions.


ServiceGroup - Tom Maguire

ServiceGroups are a useful for defining collections of related things. For
example, an
operating system contains a collection of processes.
MembershipContentRule describes entries in the group in terms of their
interfaces and
property values.
New Proposals - David Orchard

Add a binding to fully use http as a state transfer protocol: use http get
for get resource
properties, delete to delete, etc. This requires the use of WSL 2.0 web
method construct
which specifies the http method associated with a WSDL interface (because
the
application uses them directly).

Q (Chris): Why couldn't the reference property be a cookie, and where would
it come
from?
A: The cookie could be in the address field of an EPR and then gets echoed
back by the
client.

Q (Sue): Cookies are used to distinguish clients, but reference properties
are for resources
on the server side. Aren't these things differently used?
A: True, but cookies can do other things in the case where http is a
peer-peer protocol, as
it is when used for Web Services rather than Web page retrieval.

This proposal also makes the requests bookmarkable, and the responses
human-readable.
This 'webbish' view places resource id in the URI, and property id as a
secondary id.

Four things are needed: EPR without ref properties, Property id, binding
between http
and message headers, binding queries to URI.


.
Issues
Overview
1. Aggregation model - origin is lost.

Q (Fred): This will apply also when WSDL 2.0 is used?
A: Yes

3. What is returned if a property is null?
4. Get for an Any - what is the response?
6. Make description of properties consistent in the specs (either schema or
instance)
7. WS-Addressing is not a standard
8. API needed to list the properties, and allowed extensions.
10. Need meta-data on properties.
13. Should 'service' and 'port' be required in the address.

Q ( SteveG): Could we add portType, too?
A: Perhaps.

14. Can we change the name from ResourceIdentifier to Resource Id.
(=Identity)
15. Provide example for maxOccurs > 1
16. Add queryExpressionDialect.
18. Allow relative time for termination.
21. Separate set resource properties to set/update/delete.
23. Combine the retrieval ops into one op with three dialects.
25. Fault should indicate what happened in a partially successful
multiple-update op.
26. Missing invalid-value (out of range) fault in set.
27. Need a RESTful way of accessing Properties (getting them all)
28. Error code is platform-specific.
29. Discuss reconciliation between UDDI, WSIL and WS-ServiceGroups.

Q ( Fred): WSDL needs to be discussed.
A: This is an issue.

30: Not clear what the SOAPAction field should be set to
32. SubstitutionGroup makes it difficult for tools (eg to support
SetResourceProperties)

Q (Fred): WSDM proposed extensions instead.

33. How many notifications are sent when multiple properties are changed.


Discussion and New Issues list:

Q (Dave O): Can version compatibility and evolution of the specs be an
issue?
A: Noted.
Issue: Version compatibility and evolution of the specs

Q: There should be a definition of what a Stateful Resource is?
A (Ian): Yes, there is a need for such a definition - accepted as an Issue.
Issue: Definition of what a Stateful Resource is

Q (Savas): Will there documentation for the requirements?
A: There is nothing chartered, but there could be a proposal.

Issue: Requirements document is desirable. Sources, OGSI Primer/WSDM
requirements,
DAIS requirements group in GGF.

Q (Fred): Is it true that some schema constructs are not possible in
WS-Resource
properties
A: This is an issue on the issue list. Choice is not possible. Sequence is.


Issue: Consider doing a WSDL 2.0 rendering.


Anish Proposal re EPRs
References identify Web Services as targets and as message parameters.
Proposal: "An
EPR cannot be used to reference a Web Service".  The EPR reinvents
something like the
WSDL service element, but is insufficient. It contains a URI which is the
end-point for
messages, but this is not the whole story - it doesn't reference the WSDL
document.
We could use WSRef instead & create a new SOAP header block which can be
dealt with
by infrastructure. The WS-MessageDelivery spec (proposed to W3C) describes
how this
would work.

Issue: Is EPR the right construct to use for a WS-Resource Pointer.

Issue: Is there a way to separate the addressing (of a WS-Resource) from
the rest of
WSRF.

Issue: Are there valued alternatives to the implied resource pattern?

Dave Orchard Proposal:
Add a binding to fully use http as a state transfer protocol: use http get
for get resource
properties, delete to delete, etc. Needs WSDL 2.0 Web Method.
      Resloved (no spec issue), but action pending to develop binding.


Issue: Extend scope of reconciliation of WSIL/UDDI/Service group to include
WSDL
collection of ports.


TC Strategy
Rough schedule
*     Publish updated schema (OASIS namespace, patches) v Soon
*     Publish annotated drafts soon
*     Work out issues and detailed semantics next.
Sequence:
*     Resource Properties, Resource Lifetime, Base Faults, Service Groups,
Renewable
References.
Other Work
*     Primer?
o     Use GGF's OGSI Primer as source for input.

Q(Anish): couldn't the two white papers (Modelling state ) be used as
input, too?

*     Interop Fest II ?

Roles
Secretary: Tim Banks

Supporting Minute takers - for the conference calls. TBD at the first call.

Lead and supporting editors:

WS-RP:  Steve Graham, Jem Treadwell
WS-BF:  Steve Tuecke, Lily Liu
WS-RL:  Latha Srinivasan, Tim Banks.
WS-SG:  Tom Maguire, Dave Snelling
Requirements:  ??? (unresolved)

Issues Secretary: Bryan Martin