Re: XLink charmod conformance [was: Agenda for XML Core WG telcon of 2006 February 15]

Phew, this is one muddled issue.  Two or three issues actually:

1) Björn wants XLink 1.1 to require conformance to Charmod, as required 
by Charmod sectin 2 (W3C should eat its own dog food). Norm argued that 
"XLink predates the publication of Charmod and it's not clear to me that 
adding a normative statement about conformance to Charmod is within our 
charter."  But we're doing a 1.1 here, not merely a new edition of 1.0. 
    I think a statement or two could be added to XLink 1.1 to effect this.

2) Björn claims that XLink 1.1's adoption of IRIs makes it unconforming 
with C014 of Charmod, which requires use of the Reference Character 
Model.  This Model requires that "Specifications MUST define text in 
terms of Unicode characters, not bytes or glyphs." and that processing 
be defined in terms of those Unicode characters.  XLink meets that.


3) Björn says "To address this [C014 non-conformance] XLink 1.1 
implementations must either always NFC-normalize IRIs when dereferencing 
a link and this requires IRI to URI conversion, or never.".  The 
Reference Processing Model is silent about Unicode normalization, which 
was once addressed in Charmod but was farmed out to a separate spec 
which is not yet a Rec.  I fail to see how the Reference Processing 
Model can require normalization and haven't found anything from Björn 
that explains it.

4) At one point Björn points to another thread, part of the 
IRIEverywhere-27 issue on the TAG list.  This (part of the) issue does 
discuss normalization (but not Charmod conformance).  Björn seems to 
argue that IRIs should be NFC-normalized pursuant to step 1a in RFC 
3987, which applies "If the IRI is written on paper, read aloud, or 
otherwise represented as a sequence of characters independent of any 
character encoding,".  However the dominant thinking is that step 1c 
applies: "If the IRI is in a Unicode-based character encoding", which 
calls for no normalization.  It could be argued that 3987 is somewhat 
poorly written there, and does not distinguish clearly enough between 
characters as perceived by humans ("written on paper, read aloud"), and 
characters that have already been expressed in terms of the Unicode 
standard, which is certainly the case for character as seen by XLink.

Bottom line, I think we should satisfy the commenter by requiring 
conformance to Charmod (which involves saying that XLink implementations 
and XLink-conforming content must conform), but that his case about C014 
and normalization of IRIs doesn't hold water.  At least that's the best 
I could make out from following the threads, nowhere did I find an 
explicit explanation of why lack of normalization would violate C014.

-- 
François



Grosso, Paul a écrit :
> Francois,
> 
> Please let us know what you think about the
> I18N issue below by (or during) this week's
> telcon.
> 
> thanks,
> 
> paul
> 
>> -----Original Message-----
>> From: public-xml-core-wg-request@w3.org On Behalf Of Grosso, Paul
>> Sent: Monday, 2006 February 13 11:53
>> To: public-xml-core-wg@w3.org
>> Subject: Agenda for XML Core WG telcon of 2006 February 15
> 
>> 3.  XLink update.
> 
>> Re: XLink 1.1: Charmod conformance
>> ----------------------------------
>> http://lists.w3.org/Archives/Public/www-xml-linking-comments/2
> 006JanMar/0048
>> Björn says XLink 1.1 should NFC-normalize IRIs.
>>
>> Richard replied at:
>> http://lists.w3.org/Archives/Public/www-xml-linking-comments/2
> 006JanMar/0070
>> Bjoern replied at
>> http://lists.w3.org/Archives/Public/www-xml-linking-comments/2
> 006JanMar/0075
>> and Richard re-replied at
>> http://lists.w3.org/Archives/Public/www-xml-linking-comments/2
> 006JanMar/0089
>> and the thread ends there (as of the writing of this agenda).
>>
>> We still need to close this thread. *****
>>
>> We believe that our specs are correct as is, and perhaps
>> the Charmod spec could be clearer here, but that isn't
>> our remit.
>>
>> ACTION to Francois:  Review this thread and send email
>> to the XML Core WG.
>>
>> Unless Francois comes back and says we're wrong, we will
>> leave it as is and flag this issue as an outstanding dissent 
>> in the DoC, perhaps with a note that we've sent this issue
>> to I18N.
> 
> 
>

Received on Wednesday, 15 February 2006 15:55:18 UTC