Re: [xml-dev] Is Schematron (using XPath 2.0) functionally a supersetof

[Philippe Poulard <philippe.poulard@sophia.inria.fr>]

Hi Bryan,

bryan rasmussen a écrit :
>> 1) bind typed datas. Although this is not about validation but rather
>> for applications, one could imagine to involve a typed data in an
>> assertion ;
> 
> I think that is dependent on an application being written that uses
> Schematron to do datatype binding.
> 
> One of the differences between Schematron and XML Schema is that with
> XML Schema you are assured (sort of) that your 'objects' have the
> complete structure provided. Whereas in Schematron you do not have
> this assurance. I think however that in a language like JavaScript
> getting around this difficult would be easy, and in a language like C#
> it would be horrendous.
> 
>> I don't know how Schematron could take care of that ; in any
>> case, before binding typed datas, Schematron cannot defined custom typed
>> datas
> 
> http://www.oreillynet.com/xml/blog/2007/10/converting_xml_schemas_to_sche_4.html

Interesting implementation.
Unfortunately, Schematron doesn't supply typed datas. After the 
validation stage, the information is lost ; if an application has to 
parse again itself the raw data (say to get a decimal), then there is no 
advantage to express it at the validation level.
If the purpose of a schema is limited to validation, then Schematron can 
do the job ; XSLT and XQuery too.

> 
>> 3) Schematron doesn't act on content models (that is to say to what is
>> allowed to find at some place) : within an editor, one can propose an
>> element that Schematron would refuse ; for this reason, acting on
>> content models is certainly more reliable or more smart...
> 
> 
> I think that depends on the editor implementing it surely?

I guess that some editors are trying hundreds of candidate tags for 
pruning the list of what is allowed in a given context, so that the user 
will select the right thing directly ; although it's a solution that may 
work, it's an ugly solution ; I admit that there are common problems for 
which we don't care checking some assertions after the user has entered 
its input (and that would invalidate it), but there is also a class of 
problems for which there is a more elegant workaround by designing 
content models that can adapt themselves to the data to allow (and that 
are appliable for typed datas as well as for element types)

> 
> Actually also on the large 1 GB documents I guess there is an
> implementation out there, if you run the implementation through the
> DataPower XML Accelerator or something like that
> http://www-306.ibm.com/software/integration/datapower/xa35 maybe IBM
> should focus on that in their marketing. :)
> 
> Cheers,
> Bryan Rasmussen

-- 
Cordialement,

               ///
              (. .)
  --------ooO--(_)--Ooo--------
|      Philippe Poulard       |
  -----------------------------
  http://reflex.gforge.inria.fr/
        Have the RefleX !