WS Description WG

18 May 2006

See also: IRC log


[IPcaller], Arthur_Ryman, Roberto, Amelia_Lewis, Jonathan_Marsh, Allen, TonyR, Plh, Gilbert_Pilz, Youenn
TonyR, Allen




<TonyR> SCRIBE: TonyR

<Jonathan> Arthur: Completed interchange schemas for extensions, with different namespaces.

<Jonathan> ...Moved name into a base namespace

arthur: used an extra namespace to permit reuse

jonathan: find it not aesthetic, but understand the need

arthur: avoiding circular dependency

jonathan: hard drive failure lost previous "implementation"

omnes: sympathise

jonathan: producing a different version with improved design

arthur: ran new version with the new extensions: safe and signature

jonathan: roll-up report generated from test suite metadata - noted that it doesn't include Sparql, for example

john: wrote some new test cases, not had time to check them in yet.
... including nested imports

jonathan: my XSLT will fail those

arthur: xerces won't handle fragment ids correctly

lawrence: have to handle that with custom code

arthur: have a solution strategy for it
... have to use a "magic" API to handle it
... documentation for this can be found in a xerces JIRA submitted by me
... can avoid the need: anything that can be put in an in-line schema can be put in another resource

john: spec doesn't say that QNames in the extends attribute have to be unique

arthur: should be a validation error if a QName appears twice in this attribute

john: should raise this as on the list?

jonathan: yes

john: if I find any issues in Part 2 I'll raise them next call

lawrence: there is some traffic on the Woden list on some issues in implementing WSDL 2 in Axis

john: suggestion for how to integrate Woden into Axis 2 by connecting the Woden object model to their WOM
... suggestion made by Dims - may not be the most efficient, but pragmatic for the interop event
... have had a WSDL 1.1 -> WSDL 2 converter submitted to the Woden repository this week

arthur: but there are lots of things in WSDL 1.1 that cannot be represented in WSDL 2.0, so this is not a true representation of the WSDL 1.1 original
... we could define extensions to represent the missing parts of WSDL 1.1

john: add these to the spec?

arthur: no, not as part of the spec, per se, but as third party extensions to provide a migration path

jonathan: seems a bit surreal - we went to the trouble to remove all these things, and now we will add them back in!

arthur: it's for migration purposes. We're not encouraging people to use them.
... not having those features could act as a barrier to conversion to WSDL 2.0

lawrence: if we provide all these features, why would they migrate?

arthur: if we do this, then Axis 2 can become widely deployed, and people will have the ability to use WSDL 2

john: well, it would give them an easier migration path

arthur: first step would be to provide WSDL 1.1 to WSDL 2.0 conversion, as best we can. There are problems in the service element - many things you could put there that can't be represented
... you could put as many bindings as you wanted into one service element in WSDL 1.1, for example

john: the conversion tool addressed that by generating multiple services, one for each binding

arthur: may need a convention to capture where the multiple services came from

john: where would the document go?

arthur: would suggest that the document describing these extension/s be submitted as a Note

john: we can start with the work that Matt has done in this conversion tool

arthur: he wrote up some of the issues earlier

All done for this week

<Bozhong> bozhong is on the phone

<Jonathan> zakijm, who's on the phone?

<Bozhong> ipcaller is bozhong

<Allen> SCRIBE: Allen

<plh> http://www.w3.org/2002/ws/desc/6/07/interop-logistic.html


Next telcon, next week

interop report

JM: Arthur refactored all the schemas

RDF pub status

Philippe: RDF now published

JM: RDF now in last call

features and properties

JM: in a holding pattern, waiting for input


Roberto: No retionale for having an extension modify components of imported documents
... no objection to proposal

JM: How bad are the effects of the status quo?
... Is n**2 validation time really that bad?

Arthur: Yes, seeing bad performance because documents can't be cached. With restriction documents would only need to be read once.
... We are seeing workspaces with hundreds of documents

JM: Can we leave this open for one more week? Asir had some questions.


Arthur: In SOAP binding, is there any constraint between interface MEP and SOAP MEP?

Amy: SOAP 1.2 binding says that SOAP MEP uses particular WSDL MEP. Right?

Arthur: no

Roberto: maybe it is done indirectly. SOAP spec requires certain messages which correspond to WSDL MEP requirements.

Amy: SOAP 1.2 doesn't depend on WSDL. WSDL SOAP extension does depend on SOAP so that is where the restriction should go.

JM: Does the restriction exist somewhere? If so, we can fix this with a link.

Arthur: 5.10.4 binds SOAP 1.2 MEPs to WSDL MEPs.

JM: Is this complete?
... Just talks about WSDL in/out MEP
... We need someone to make a proposal to add this information

<scribe> ACTION: Amy will write a proposal for CR029 [recorded in http://www.w3.org/2006/05/18-ws-desc-minutes.html#action01]


Arthur: Do you require a default value for SOAP MEP
... Add some rule for computing SOAP MEP default so it is not really optional

Amy: no

Arthur: Is it required that there be a SOAP MEP assigned to every operation in a binding?

JM: 5.10.3, second bullet defines this - defaults to request/response

Amy: When oneway MEP is introduced this will break

JM: It will make the attribute required to handle this case

Arthur: can add a link to this section as an editorial item

Amy: move that SOAP MEP default or SOAP MEP must appear

Arthur: wants the link in any case

JM: amend Amy's proposal to include the link
... Any objections to adding the link in any case?

Resolved: add a link to 5.10.3

Amy: change from having default of WSDL in/out to being an error if you don't specify SOAP MEP default or SOAP MEP

Resolved: CR030 closed with addition of forward reference


Arthur: no two components can have the same ref property

JM: only way to get this is to have more than one sibling soap modules that have different required values

Arthur: or to have multiple instances of the same module
... should just be an error
... validator should say it is wrong. Processor can do whatever it wants.

Roberto: not convinced by this because of extensibility. Someone chould add parameters to module with extensions. May require them to write more than one soap module.

Amy: only one scope for soap modules so you can only have one.

Robero: so it is easier to understand that there is a single soap module with multiple parameters stored underneath.

Arthur: soap module is an element so there is room to put whatever you want underneath.

Roberto: withdraws objection to proposal.

Resolved: add constraint to 5.8.2 that a soap module component on a given binding, ref will be constrained to be unique.


JM: should be a constraint on soap headers that each soap header should be uniquely identified

Arthur: is there a one to one correspondence between a header block and a header on the wire?
... do child elements of a header all have unique element names?

Roberto: no

Arthur: when you put in multiple header blocks, can some have mustunderstand and others not?

JM: yes but it won't change meaning
... don't see a compelling reason to change this

Roberto: it would be more understandable if we allow only one soap header for a given element name. If we want to have multiple headers add an attribute like repeatable.

JM: currently we can specify an exact number, not just many

Arthur: we can now say something like have between 3 and 5 headers with minoccurs and maxoccurs.

JM: proposal - change 0 or 1 to 0 or more and require each header block to have a unique declaration.

Roberto: status quo doesn't allow any number of headers of the same type.

Arthur: proposal - require header element declaration to be unique. If you want to put in more than one header define a complex type.

JM: proposal - element declaration is unique but before we close, ping Glen and Dave O

<Jonathan> ACTION: Marsh to ping DaveO and Glen about the proposed resolution to CR032. [recorded in http://www.w3.org/2006/05/18-ws-desc-minutes.html#action02]


<Jonathan> "namespace#name" -> "qname"

Resolution: change component identifier from namespace #name to QName


JM: editorial - remove MUST NOT

Resolved: CR034 closed


Arthur: no default http method. Suggest POST.

JM: if we add safety then if tagged as safe you get GET otherwise POST.

Resolved: close CR035 by adding "otherwise the value POST" as last bullet.

<Jonathan> Proposal: add a fourth bullet to 6.3.1; "Otherwise, the value 'POST'".

Summary of Action Items

[NEW] ACTION: Amy will write a proposal for CR029 [recorded in http://www.w3.org/2006/05/18-ws-desc-minutes.html#action01]
[NEW] ACTION: Marsh to ping DaveO and Glen about the proposed resolution to CR032. [recorded in http://www.w3.org/2006/05/18-ws-desc-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.127 (CVS log)
$Date: 2006/05/18 16:36:44 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.127  of Date: 2005/08/16 15:12:03  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/service elements/services/
Found Scribe: TonyR
Inferring ScribeNick: TonyR
Found Scribe: Allen
Inferring ScribeNick: Allen
Scribes: TonyR, Allen
ScribeNicks: TonyR, Allen

WARNING: Replacing list of attendees.
Old list: Roland_Merrick TonyR jkaputin Arthur_Ryman Jonathan_Marsh +1.905.413.aaaa lmandel
New list: [IPcaller] Arthur_Ryman Roberto Amelia_Lewis Jonathan_Marsh Allen TonyR Plh Gilbert_Pilz Youenn

Default Present: [IPcaller], Arthur_Ryman, Roberto, Amelia_Lewis, Jonathan_Marsh, Allen, TonyR, Plh, Gilbert_Pilz, Youenn
Present: [IPcaller] Arthur_Ryman Roberto Amelia_Lewis Jonathan_Marsh Allen TonyR Plh Gilbert_Pilz Youenn
Got date from IRC log name: 18 May 2006
Guessing minutes URL: http://www.w3.org/2006/05/18-ws-desc-minutes.html
People with action items: amy marsh

[End of scribe.perl diagnostic output]