Re: remove xml-stylesheet appendix?

One factoid that may be relevant:

A grddl implementation that automatically applies stylesheet PIs will 
fail the test suite, fairly badly: the library functions both use a 
stylesheet PI for presentation in an HTML user agent. A GRDDL aware 
agent needs to ignore these to get the XSL.

I suspect this means that the exact wording is less important, since an 
implementer with a broken XSL/XML implementation is likely to have that 
thrust in their face.

Jeremy

Dan Connolly wrote:
> On Fri, 2007-04-06 at 09:57 -0400, Chimezie Ogbuji wrote:
>> I would be against removing it.  Although they are not enforced
>> 'formally', xml-stylesheet processing instructions are supported by a
>> large number of web agents and XML processors.  In addition, I think it
>> is in the interest of GRDDL to be clear about how its use of XML
>> pipleines to produce content for machine consumption (faithful
>> rendition) differs significantly from the precedent of using XML
>> pipelines to produce content for human consumption (the whole content
>> versus presentation pattern: docbook -> PDF,HTML, etc..).
>>
>> GRDDL aware-agents which piggy-back XML processors may be in for a
>> surprise if the underlying XML processor, applies xml-stylesheet
>> processing instructions by default (some do).  An explicit health
>> warning is prudent, even if we explicitly mark it as informative.
> 
> Then the health warning should say that GRDDL-aware agents should
> ignore stylesheet PIs, no?
> 
> It seems to me that mixing stylesheet PIs and GRDDL links
> freely should work fine. I'm not interested to advise
> authors to avoid it.
> 
> Sorry for the delay in responding.
> 
>> I'd prefer clarifying the text rather than deleting it or marking it as
>> informative (I assumed that its place in the appendix suggests that is
>> informative).  A new first paragraph:
>>
>> [[[
>> The xml-stylesheet processing instruction[STYPI] is generally deployed
>> for automated presentation processing. This type of link is different
>> from links to GRDDL transformation algorithms, which are intended to
>> facilitate the extraction of RDF as a faithful rendition [#sec_rend] of
>> the source.  The former is geared more for human consumption while the
>> latter is primarily for machine consumption. 
>>
>> Document authors who wish their documents to be unambiguous when used
>> with GRDDL should avoid using xml-stylesheet processing instructions as
>> their use may interfere with transforms nominated by GRDDL for the
>> production of GRDDL results in the same source document.
>> ]]]
>>
>> The last part of the above was added with language that ended up having
>> to contend with statements made in the faithful-infoset sections and the
>> new health warnings added about DTD's and entities, so I tried to keep
>> the tone consistent.  
>>
>> The second paragraph (from the original) is not true as there are at
>> least three examples of XSLT processors which support this: 4Suite,
>> Saxon, and MSXML (I'm not sure why this did not come to my attention
>> before).  
>>
>> [[[
>> Also, parsing the content of processing instructions is not supported by
>> XML tools such as XSLT processors, and grounding processing instructions
>> in URI space is not as straightforward as using namespaces with
>> attributes.
>> ]]]
>>
>>
>> I'm not sure can suggest a change for that second part, so my vote is to
>> delete it.  Dave's concern seems to be about wandering into
>> implementation advice.  I don't think a health warning about possible
>> clashing of transform nomination WRT to xml-stylesheet applies, but a
>> (false) statement about support of xml-stylesheet along with a critique
>> of them doesn't seem very appropriate, in retrospect.  Especially
>> considering the language of the failthful-infoset paragraph suggests the
>> WG has taken a stance of being silent about XML processors.
>>
>>

-- 
Hewlett-Packard Limited
registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Tuesday, 24 April 2007 09:50:18 UTC