RE: Proposal for Issue 6700

I'm copying the relevant conformance section of XML Infoset spec [1]
here for discussion:

 


3. Conformance


Since the purpose of the Information Set is to provide a set of
definitions, conformance is a property of specifications that use those
definitions, rather than of implementations. 

Specifications referring to the Infoset must: 

*	Indicate the information items and properties that are needed to
implement the specification. (This indirectly imposes conformance
requirements on processors used to implement the specification.) 
*	Specify how other information items and properties are treated
(for example, they might be passed through unchanged). 
*	Note any information required from an XML document that is not
defined by the Infoset. 
*	Note any difference in the use of terms defined by the Infoset
(this should be avoided). 

If a specification allows the construction of an infoset that has
inconsistencies as described above under Synthetic Infosets
<http://www.w3.org/TR/xml-infoset/#intro.synthetic#intro.synthetic>  it
may describe how those inconsistencies are to be resolved, and should do
so if it provides for serialization of the infoset. 

Thanks.

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

 

Li

________________________________

From: Gilbert Pilz [mailto:gilbert.pilz@oracle.com] 
Sent: Tuesday, September 01, 2009 1:34 PM
To: Ram Jeyaraman
Cc: Li, Li (Li); public-ws-resource-access@w3.org;
ashok.malhotra@oracle.com
Subject: Re: Proposal for Issue 6700

 

Comments inline . . .

On 9/1/2009 10:07 AM, Ram Jeyaraman wrote: 

My overarching point / observation is that we are washing our hands of
doing the actual XML Infoset conversion work and asking the reader to do
the work if she/he cares about it. I doubt if anyone would go to the
length of doing the conversion and even so such a conversion is
non-normative and cannot be considered a part of the specification.

You are correct. That is the point of this approach. We want a way to
allow serializations other than XML 1.0 to be considered as conforming
without making the 99% of the readers who don't care about such
serializations wade through Infoset descriptions of each element and
attribute. I'm not sure what you mean by "such a conversion is
non-normative". If there is a one-to-one mapping between XML and
Infoset, which terminology we use is, in the abstract, somewhat
arbitrary. We are choosing to use XML terminology because we expect that
the majority of our readers will be more comfortable with this
terminology.



Having said that, I believe that the XML Infoset specification defines
how to map from XML to the Infoset representation. Are we inventing
anything new in the mapping rules that is not already specified in the
XML Infoset specifications? Why can't the WS-RA specifications simply
refer to the XML Infoset specification for the mapping rules?

You make a good point. Let's discuss on the call today . . .



 

Thanks.

 

From: Gilbert Pilz [mailto:gilbert.pilz@oracle.com] 
Sent: Friday, August 28, 2009 11:59 AM
To: Ram Jeyaraman
Cc: Li, Li (Li); public-ws-resource-access@w3.org;
ashok.malhotra@oracle.com
Subject: Re: Proposal for Issue 6700

 

I think the purpose of doing it this way is twofold:

1.) It saves the effort of having to go through the description of every
element and attribute and change that description to Infoset
terminology.

2.) It reduces the potential for confusion inherent in the use of
Infoset terminology. Like it or not, 99% of developers use XML 1.0
exclusively and tend to think in XML 1.0 terms. Requiring them to map
between Infoset terminology and XML 1.0 terminology makes the spec
harder to read.

I think you are confusing "defined in" with "described by". By asserting
that the specifications are "defined in" Infoset we are declaring that
the normative definition of the elements and attributes in the spec are
based on a abstract document structure. This allows for implementations
that may use serializations other than XML 1.0. The fact that these
elements and attributes are "described in" XML 1.0 terms is merely a
convenient, notional shorthand that makes it easier for 99% of the
readers to understand. As long as the mapping from XML 1.0 to Infoset it
complete, there shouldn't be any problems.

Which brings me to my next point, how is the mapping 'partial'?

- gp

On 8/27/2009 5:42 PM, Ram Jeyaraman wrote: 

I like to question the purpose of this exercise (issue).
 
  

	This specification is defined in terms of XML Information Set
(Infoset) and not in terms of XML 1.0, even though the specification
uses XML 1.0
	    

terminology.
 
The first part says that the specification is defined in Infoset
notation, but the later part says that the specification is uses XML 1.0
terminology. Is it true that the specification is defined in terms of
Infoset?
 
Further, the mapping provided in the second paragraph of the proposal is
a partial mapping and not a complete one.
 
What is the point of leaving the exercise of mapping the XML document to
Infoset as an exercise to the reader. What do we gain by saying this?
 
-----Original Message-----
From: public-ws-resource-access-request@w3.org
[mailto:public-ws-resource-access-request@w3.org] On Behalf Of Li, Li
(Li)
Sent: Tuesday, August 25, 2009 7:47 AM
To: public-ws-resource-access@w3.org
Cc: ashok.malhotra@oracle.com
Subject: Proposal for Issue 6700
 
 
This proposal is based on the consensus between Ashok, Wu and Gil. To
complete the Infoset description for WS-Eventing, we propose to just add
the following text at the end of "Section 3.1 Notational Conventions" of
WS-Eventing, instead of following the approach taken by 6424:
 
---
This specification is defined in terms of XML Information Set (Infoset)
and not in terms of XML 1.0, even though the specification uses XML 1.0
terminology. A mapping from XML to Infoset is straightforward as
described below, and it is recommended that this should be used for any
non-XML serializations.
 
XML documents map to Infoset Document Information Items.  XML Elements
map to Infoset Element Information Items (EIIs) and attributes map to
Infoset Attribute Information Items (AIIs).  The *children* property of
an Element Information Item (EII) is a collection of EIIs corresponding
to its Element Children.  The *attributes* property of an EII is a
collection of AIIs corresponding to its attributes.
 
See the Infoset specification [http://www.w3.org/TR/xml-infoset/] for
more details.
---
 
We hope this proposal can be applied to related issues 6701-6704 as
well, by adding the above text to the "Notational Conventions" section
of each corresponding spec.
 
Li Li