Re: new issue? dfxp and language selection

Hello TT group,

On 4 Dec 2008, at 13:10, John Birch wrote:
> For me the semantic of xml:lang in our specification is only ever  
> informative about the element content - not the document content. It  
> is an attribute value on which a processor might make a selection,  
> in the same way that an xml document representing a database of  
> addresses might be processed by a processor that uses an attribute  
> value to 'select' only addresses in a specific geographic region...  
> The choice of which specific attribute instances should be used for  
> a processing decision is up to the processing - not the document  
> structure designers.


I disagree.

The problem in the analogy you are making is that database of  
addresses does not exhibit any behavior: in this case XML is merely  
used as a storage mechanism using a well-defined encoding method (i.e.  
the database schema).

TimedText DFXP, like SMIL, exposes presentational and behavioral  
semantics for which the processing conformance has to be clearly  
defined. In a distribution format, if the intent is to perform content  
selection (e.g. based on the xml:lang attribute), then this has to be  
clearly specified. In an authoring format (e.g. AFXP), I would be more  
inclined to allow various kinds of processing (i.e. application- 
specific behaviours).

In SMIL, the ContentControl module provides the "switch" element to  
enable user-agents to perform content selection (pruning data either  
on-the-fly or during pre-processing stage), but the selection/test  
attributes can also be used inline, without the "switch" container.  
This selection behavior is specified for a well-defined set of test  
attributes, because the intent has to be extremely clear to user-agent  
implementors.

http://www.w3.org/TR/2008/REC-SMIL3-20081201/smil-content.html#ContentControlNS-InlineTestAttributes

I think that xml:lang in TT/DFXP should not be used for content  
selection, but merely to indicate the content language as per the XML  
1.1 recommendation. Content selection introduces complex issues such  
as the logical combinations of the selection criteria (e.g. [en-US OR  
en-GB] versus [en-US]), for which SMIL has a limited solution based on  
coma-separated values:

http://www.w3.org/TR/2008/REC-SMIL3-20081201/smil-content.html#adef-systemLanguage

Food for thoughts.

Kind regards, Daniel.

Received on Thursday, 4 December 2008 13:47:26 UTC