Minutes PFTF 20030225

Participants:
 Jean-Jacques (scribe), Amy, Philippe, Glen, Steve, Colleen, Don, David,
 Sandeep
Regrets:
 Sanjiva, Paco, Youenn
Missing:
 Jonathan, Chris

Previous minutes:
 http://lists.w3.org/Archives/Public/public-ws-pnf-tf/2003Feb/0047.html
Agenda:
 http://lists.w3.org/Archives/Public/public-ws-pnf-tf/2003Feb/0051.html

Glen: extra 30 minutes?
 (no objection)

Glen:

1) Indicate a particular feature is provided
2) Indicate a particular feature is required
3) Indicate a particular property constraint

 Considering that each of these things may be scoped in various ways
 (service / portType / operation / binding), does that about cover what
 we need to do here in WSDL?

Glen: anything else missing? the list should enable us to quickly 
evaluate solutions.

Don: sounds good; but the group will also be interested in 
knowing why the standard extension is not enough.

Glen: agree.

Don: pick one feature, do it side by side

Amy: lower barrier and provide modularized property set

Glen: like amy's example, however properties critical for implementing,
not describing

Amy: disagrees, you find it in the choreography spec (WSCI, BPEL)

Glen: they define a slot for correlation id. they wouldn't say
correlation id is "5", woudln't make sense. content id or web method
example might be better

don: the better the demonstration the better the example

Glen: no WSDL binding today to declare using SOAP-Response MEP and
Web-Method feature set to GET

Amy: look also at scenario 5, to be less SOAP centric?

Glen: this particular property goes into this particular MIME header?

Amy: yes, essentially

Amy: SOAP and HTTP would in my view restrict the discussion to SOAP

Glen: however, SOAP and HTTP well understood

Amy: agree, as long as not restricted to SOAP and HTTP

Glen: are people comfortable with expressing features at abstract level?
should show this through examples

Don: simple and easy to understand examples

Glen: have wide range of scenarios already. some of them require WSDL
descriptions. proposal to dive in and to that?

Glen: first class feature or property attribute. S7 interesting, shows
given propertyConstraint affects more than one feature. several proposal
suggested already. maybe do by email, rather than IRC

don: agree

Glen: so, now, go into the scenarios in detail and offline propose
syntax. S2 good example. however, too complex. could reuse my earlier
security example (not certificates, just raw HTTP authentication)

sandeep: [...]

Glen: anyone having problem with feature at the abstract level?

[assent]

Glen: abtract security feature. implemented by concrete feature either
through module or underlying protocol

Sandeep: so would provide secure channel

glen: yes

Sandeep: abstract feature would be expressed by namespace, name/value
pair. how it gets mapped would be independent

Glen: yes. properties in SOAP 1.2 used to be qnames. however, now they
are URIs. so, could do same thing, but syntax would be different. some
fairly good suggestions have been made regarding specifying property
constraints. unanswered question: where can property constraints go?
e.g. you might want to set the value of a property just for the inbound
message

Sandeep: might also want to restrict value to a string of a certain
length

Glen: don't think property constraints should be within features, since
properties might be shared

Amy: suggest have constraint attribute and value attribute, which would
contain either value or token list

Glen: understand optimisation; however, two separate issues

Amy: probably right

Glen: that deals with typing. where to they go? one container element to
contain properties (whether at service, portType, etc. level).
properties are somewhat orthogonal to features (can be shared by
features). non-repudation feature would refer to a user name, which
would be a property defined by an authentication feature

Amy: how would you refer to an element?

Glen: if using WSDL extensibility?

Amy: no, if want to refer to property... one property set to the value
of another property (probably advanced function). qnames as property
names are analog to pointers in programming languages, not URI

David: why?

Amy: URIs have been abused

Glen: infoset value space of particular xml tree; virtual URI space of
properties, already defined by SOAP. URIs have benefit of making
properties compliant with semantic Web (policy assertions, etc.)

David: and policy assertions
 [David's cat agrees]

Don: if can share properties between features, would you need to scope
properties to features?

Amy: no

Glen: up to the spec writer

David: there's a corresponding TAG finding

Don: if property has same URI and default value, and want to override
default in second feature only?

Glen: cannot do this. properties must be defined uniquely

David: wonderfull description; need to see it written

Glen: need to test SOAP 1.2 endpoint on inbound flight. maybe on way
back

David: could volunteer, but would be difficult to recreate your nice
wording

Glen:

1) crisp definition of choosen scenario
2) syntax proposal by Thursday EOB PST
3) select best proposal by Friday EOB PST

Glen: one or two examples?

David: 1

Amy: 2

David: 2 if enough volunteers

Glen: 1st example: security example

Amy: if 1 example only, need properties

Glen: would then suggest Web Method feature

Amy: hum... don't understand WebMethod

Don: might have pushback; why not done by SOAP binding itself?

Amy: share concern

Glen: well, done generically, since might want to tweak the abstract
feature

Jean-Jacques: maybe also use WebMethod to set the method for HTTP
binding, not just SOAP binding

Glen: yes, could set GET at abstract level, and HTTP binding would infer
that the HTTP method is GET

David: follow the REST guidelines?

David: would separate Web Method from REST

Glen: agree, not talking about REST

Glen: not sure the WSDL extensibility mechanism would suffice. talking
about framework around extensibilty mechanism. akin to SOAP enveloppe

David: it's a framework too

Glen: shall we go for this one as well?

Jean-Jacques: sounds good to me

Glen: need 3 volunteers. description with a) extensibility and b)
properties and features. volunteer to write up the scenarios. need
volunteers to grab each one and propose syntax

Jean-Jacques: could spend some time but what to do with extensibility?

Glen: could do:
 <soap:binding>
   <webmethod>GET</webmethod>
 </soap:binding>


Glen: anyone else?

Amy: don't quite like what I know of WebMethod. might do security stuff;
but need starting point

Glen: will provide starting point. continue by email

ACTION: Glen to decribe the scenarios we're focusing on
ACTION: Jean-Jacques to help out with writing them up
ACTION: Glen to make dinner reservations for Sunday night

Received on Wednesday, 26 February 2003 12:01:03 UTC