RE: updated editor's copy of WSDL 1.2 spec

Take a look at the XML Schema spec[1]. I intend that the abstract model
for WSDL will be along similar lines. I will be working on it this week,
so expect that section of the spec to change drastically

Gudge

[1] http://www.w3.org/TR/xmlschema-1

-----Original Message-----
From: Jean-Jacques Moreau [mailto:moreau@crf.canon.fr] 
Sent: 24 June 2002 09:26
To: Sanjiva Weerawarana
Cc: WS-Desc WG (Public)
Subject: Re: updated editor's copy of WSDL 1.2 spec



Hi Sanjiva,

My initial reaction is to say no, the abstract model should not be
coupled to the infoset. But then I am wondering what does this really
means. Is the difference only in terms of terminology ("property" vs.
EII?) or is it more profound? Wouldn't both approaches essentially model
a (DOM) tree? Isn't the infoset already a suitable model?

The cut we have done for SOAP 1.2 is to describe the
semantics/processing [1] separate from the syntax [2]. Would a similar
model work for WSDL?

Taking a specific example from your latest draft -section 2.2 [3]-,
would it work to keep to keep only paragraph 1 and move the rest to
section [3], whilst adding a longer description of what a message
represents?

I realize I am raising more issues than providing answers... What do you
think?

Jean-Jacques.

[1]
http://www.w3.org/2000/xp/Group/1/10/11/soap12-part1.html#msgexchngmdl
[2] http://www.w3.org/2000/xp/Group/1/10/11/soap12-part1.html#soapenv
[3]
http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/part1/part1.html#messag
e-desc-component



Sanjiva Weerawarana wrote:

> <snip/> I was wondering where the
> semantics go .. in the abstract description or at the point of 
> describing the infoset for each description component?
>
> I wonder whether we should drop the "be infoset based" requirement now

> that we have are abstract model based. I kind of like the infoset 
> description approach (I cut-n-pasted from the soap spec to get the 
> template; thanks to whoever wrote that part!), but it does seem a bit 
> redundant.
>
> <snip/>
> >    * Re. "property". Shouldn't this be EII or AII in a number of 
> > places?
>
> I didn't think the abstract model should be coupled to do the infoset.

> Do you? EII/AII implies a specific serialization .. one can imagine 
> more than one serialization (infosets) of the same abstract model.

Received on Monday, 24 June 2002 06:16:39 UTC