Re: [CDR Framework] i18n comment: Language identification for child documents

Bjoern Hoehrmann a écrit :
> It seems what you want is that the document element in SVG documents
> inherits language information from referencing elements if no other
> language information is available. So if you have a graphic without
> any language information referenced from a german XHTML document, the
> graphic should be considered a german graphic.

Yes.  The scenario I have in mind is the following: I have an XHTML doc 
containing an SVG image.  There is an xml:lang="de" somewhere in the 
XHTML, which applies to the SVG image by virtue of being "above" it in 
the tree; there is no xml:lang in the SVG.  And there is something 
observable (styling, behaviour, whatever) in the image or in script that 
depends on language.  In this state of affairs the language is German 
and I observe proper German styling/behaviour/whatever.

Now I edit the document and externalize the image (perhaps because I 
reuse it several times in the document). It is my strong sentiment that 
the mere fact of externalizing the image should not change its styling 
or behaviour, which would happen if the language information is lost in 
the process.

The CDR framework seems to go to quite some length to make sure that the 
separate DOMs can communicate and that e.g. events propagate properly 
after externalization.  What we are requesting here is that this also be 
the case for the basic XML inheritable attributes (of course we i18n 
folks are mostly interested in xml:lang, but in principle it is the same 
for xml:space).

Externalizing the image does add a new possible authoritative source of 
language information: the HTTP Content-Language header.  So be it.  If 
there is authoritative information, use it; but if there is none, then I 
think that the xml:lang that applied before externalization should still 
apply after.  There is nothing new about having multiple sources of 
language information with different priorities, see for instance 8.1.2 
in the HTML 4 spec.


> If that's desired here, could you elaborate whether this is specific
> to SVG or should apply to other formats aswell

Nothing specific to (X)HTML or SVG here, it's just the first example 
that came to mind.  Anything XML has xml:lang, and HTML has lang.  Since 
CDR seems to be heavily DOM-based, I would say "anything that can have a 
DOM".


> Could you also elaborate on why such a requirement or recommendation
> would be better than simply to allow implementations to apply heuristics
> (which may or may not include what you propose) to determine the
> language?

"allow implementations to apply heuristics" sounds like a blank check to 
me, and antinomic to interoperability.  You don't want CDR browsers to 
behave randomly, do you?

Regards,

-- 
François Yergeau

Received on Thursday, 9 February 2006 22:25:25 UTC