RE: id is Candidate Recommendation (Call for Implementations)

Should it turn out to be unpalatable to move xml:id to 
a separate namespace on such short notice, a second
possibility for resolution occurred to me yesterday evening.

The intentional precedent set by the signature working group
was to leave the door open for any number of canonicalizers
going forward.  Moreover, when the namespace inheritance rules
of C14N proved to be unpalatable in a class of applications,
the response of the working group was to define a second
canonicalizer, exclusive C14N.  This algorithm is defined as
C14N plus some extras to help cease or modulate the undesired
property.

Now we are having a discussion about how some other inheritance
rule behaves a new class of applications (those that use xml:id).
Is it then simply time for another C14N algorithm?  The new
algorithm would be C14N except for ceasing or modulating the
undesired inheritance behavior (e.g. provide a list of xml 
namespaced attributes for which the inheritance should occur,
where the default is {lang,space}).

Either way, a separate issue emerges.  Some group at the core 
of the W3C needs to come to a decision about what a namespace 
means and whether additions, deletions or changes to the schema 
(the collection of names) necessitates a change of namespace URI.  
My read of the definition of namespace (a collection of names 
*identified* by a URI) suggests that an answer is yes because 
two different collections of names are not identical.

A 'no' decision (i.e. that changing the vocabulary does not require
a change of namespace URI) may have significant policy ramifications.  
For example, would it be permissible for the XForms working group 
to issue XForms 1.1 without changing the namespace URI from the one 
used in XForms 1.0?  It is very easy for me to illustrate the 
disastrous consequences of such a decision.  But what's important 
to me is that I think the current W3C policy sets the precedent 
because it doesn't permit a new version of an XML language to become 
a recommendation without changing the namespace URI relative to the
prior version.

On the other hand, a 'yes' decision implies the capacity to say
"this software is a processor for namespace X" which I've heard
is a contentious point (some say yes, some say no, but most who
say anything do it with conviction).

Best regards,
John Boyer


-----Original Message-----
From: Robin Berjon [mailto:robin.berjon@expway.fr]
Sent: Wednesday, February 09, 2005 2:02 AM
To: Martin Duerst
Cc: John Boyer; W3C Advisory Committee; W3C Working Group Chairs
Subject: Re: id is Candidate Recommendation (Call for Implementations)


Martin Duerst wrote:
>  Norm was saying something about
> canonicalization, but it would be helpful to have this
> explicitly stated.

The problem is with the way attributes in the xml:* namespace are 
inherited in C14N. You can see the root of the thread there:

   http://lists.w3.org/Archives/Public/public-xml-id/2005Jan/0037.html


-- 
Robin Berjon
   Research Scientist
   Expway, http://expway.com/

Received on Wednesday, 9 February 2005 18:25:50 UTC