Re: Types re-write for WSDL Part 1

I dislike butting into a conversation which seems primarily between the editors, but this seems to me to raise a question of process, and as it is my labor that seems likely to be disregarded, I suppose it is up to me to register at least an informal question.

At the request of Johnathon Marsh and the rest of the Working Group, I took on the task of modifying the materials related to schema languages/type systems.  This material was sent to Gudge, who suggested some modifications.  It was then published to the list.  It caused relatively little discussion, and at a subsequent teleconference, it was approved for inclusion, and Gudge was asked to incorporate the submitted text with some minor changes.

Since then, it appears that another editor, Sanjiva, has taken on the task of writing the material anew.  I don't quite understand.  Was the issue reopened at a teleconference that I missed?  I hope that that's the case, as otherwise, it seems to me that the process we have in place suggests that the submitted and approved drafts should first be incorporated, and then (if need arises) challenged and changed.

I hope that I haven't misinterpreted, particularly by being out of the loop on a conference call, and I hope that I do not offend by raising the question of process for the handling of such solicited modifications.

Amy!
On Sat, 18 Jan 2003 11:15:39 -0500
Sanjiva Weerawarana <sanjiva@us.ibm.com> wrote:

> 
> 
> 
> 
> Hi Gudge,
> 
> My understanding of the spirit of the decision we made was to support any
> type system in the component model but that there would be built-in support
> for the XSD type system in addition. If the minutes aren't clear about it
> we should ask the group rather than going with your memory vs. mine.
> 
> Why do QName references break if we don't have two separate properties? I
> cannot grok that; the single types component would have xsdTypes and
> xsdElements properties which are what the QNames refer to. So I don't see
> why the approach I was taking doesn't work.
> 
> Let me finish what I was writing and then let's go back to the group and
> decide. I suggest we go ahead and finish this draft and publish with that
> structure and change it if needed. Given the nearly 100% change from the
> previous draft it'll hardly be a problem to change something so small if
> the group agrees.
> 
> I got home a few hrs ago- I will try to work on the early tomorrow morning
> my time (which will be Saturday night US time) so that we can get this
> wrapped up.
> 
> Sanjiva.
> 
> 
> "Martin Gudgin" <mgudgin@microsoft.com> on 01/17/2003 07:45:15 PM
> 
> To:    Sanjiva Weerawarana/Watson/IBM@IBMUS
> cc:    "W3C Public Archive" <www-archive@w3.org>, "Jean-Jacques Moreau"
>        <moreau@crf.canon.fr>, <roberto.chinnici@sun.com>, "Jeffrey
>        Schlimmer" <jeffsch@windows.microsoft.com>, "Amelia A. Lewis"
>        <alewis@tibco.com>, "Jonathan Marsh" <jmarsh@microsoft.com>
> Subject:    RE: Types re-write for WSDL Part 1
> 
> 
> 
> Sanjiva,
> 
> I have not undone anything that you have done.
> 
> When did we decide that XML type systems would be treated the same as
> everything else? I think we agreed to have XML Schema as a
> fully-supported type system and to detail how one might use other XML
> based type systems ( which Amy did ) and also to allow other type
> systems to play ( by basically using similar mechanisms to those used by
> 'other' XML type systems ).
> 
> At the FTF in VA we agreed that extension elements/attributes ( which
> other type systems would have to be ) would annotate the existing
> component model with additional properties, as necessary. There was no
> suggestions that we would rid ourselves of {element declarations} or
> {type definitions} as far as I remember, nor can I find anything in the
> minutes. In fact, it stops QName references from working if we don't
> have two separate properties. I really think that having a types
> component with a single bag of stuff is the wrong approach. I suggest
> that other type systems add properties to the definitions component just
> like {element declarations} and {type defintions} and defined how
> referencing works between the attributes they specify for parts and
> these properties ( this is, AFAICT what Amy has done for RelaxNG and
> DTDs )
> 
> Either way we can't publish today as the types sections are now
> incomplete.
> 
> Gudge
> 
> 
> > -----Original Message-----
> > From: Sanjiva Weerawarana [mailto:sanjiva@us.ibm.com]
> > Sent: 17 January 2003 19:57
> > To: Martin Gudgin
> > Cc: W3C Public Archive; Jean-Jacques Moreau;
> > roberto.chinnici@sun.com; Jeffrey Schlimmer; Amelia A. Lewis
> > Subject: Re: Types re-write for WSDL Part 1
> > Importance: Low
> >
> >
> >
> >
> >
> >
> > Hi Gudge,
> >
> > I disagree - the way you had written the component model
> > there were fundamental properties called element and type
> > declarations. While those make sense for XML Schema, they
> > make no sense at all for non-XML type systems. What we agreed
> > to was that WSDL would cleanly support multiple type systems,
> > including non-XML ones. The approach you're suggesting treats
> > XML centric type systems in a first-clas maner and everything
> > else in a different way. That's not acceptable.
> >
> > The re-write I started was to make it clear that any type
> > system has the same place in the model. Have you un-done what
> > I started working on?
> >
> > (My CVS access is still busted and I'm currently in Heathrow;
> > so poor net
> > access.)
> >
> > Sanjiva.
> >
> >
> >
> > "Martin Gudgin" <mgudgin@microsoft.com> on 01/17/2003 10:35:39 AM
> >
> > To:    "W3C Public Archive" <www-archive@w3.org>,
> > "Jean-Jacques Moreau"
> >        <moreau@crf.canon.fr>, <roberto.chinnici@sun.com>, Sanjiva
> >        Weerawarana/Watson/IBM@IBMUS, "Jeffrey Schlimmer"
> >        <jeffsch@windows.microsoft.com>
> > cc:    "Amelia A. Lewis" <alewis@tibco.com>
> > Subject:    Types re-write for WSDL Part 1
> >
> >
> >
> > I've been looking at the initial start of the types re-write.
> > I think that types should remain in a separate section. I'm
> > not convinced that there is a types component at the abstract
> > level. I think we just stick with {element declarations} and
> > {type definitions} properties in the definitions component.
> > We agreed in VA that additional type systems would add their
> > own properties to existing components in the abstract model.
> > I believe that the other XML based type systems should
> > populate these properties. We then define how the QName
> > reference stuff works just once.
> >
> > Other type systems that are not XML based would need to
> > specify what properties they add to the definitions
> > components. I've added text saying that the component model
> > can be added to by extensbility elements and/or attributes.
> >
> >  Gudge
> >
>  >
> 


-- 
Amelia A. Lewis
Architect, TIBCO/Extensibility, Inc.
alewis@tibco.com

Received on Monday, 20 January 2003 10:01:55 UTC