RE: Xinclude test case

Hi Harry,

> However - can we *guarantee* this? I ran Xalan on the test 
> case and it does *not* by default resolve the XInclude, so 
> any programmer using Xalan to implement GRDDL would fail on 
> this test case.

That would only be true if it were not possible to configure Xalan to do
the Xincludes.  In which case - don't use Xalan.

> Danny - tell me when your implementation is 
> ready and I'll bet we'll get another result on this one :)
> 
> This puts the entire test case in jeopardy.

I'm not following you there.  What is the problem?

> Either we
> 
> 1) Dictate XInclude must be supported by all GRDDL parsers
> 
> or
> 
> 2) We don't, and therefore say that if you want to make sure 
> your XIncludes get resolved, use XMLProc or another Pipeline 
> implementation.

I'm not sure what you mean there.  Who is the 'you' in that sentence?
The implementor?  Its not up to the implementor.  I tend to think in
terms of the transform author.  What can a transform author assume about
a correct GRDDL implementation.

In those terms, a question is (roughly) what exactly is the input
document that transforms apply to.  The current GRDDL spec is defined in
terms on XPATH expressions and looking at the XPATH spec [1] it states
that an XML document "may be validated against one or more schemas" (see
[1] 2.2.1) producing a PSVI.  So as well as specifying what to do about
Xincludes, do we also have to consider schema validation?  What about
DTDs?  They can define default attributes on elements?

> Otherwise you have to hope for the best.

Eh?

> 
> I'll take this up as least as a straw poll on our telecon tomorrow.

I just took a quick skim through the xinclude spec [2].  In 1.2 it
states:

[[
 XInclude allows the author to provide default content that will be used
if the remote resource cannot be loaded.
]]

That set my antenae twitching.  It means, I think, that if we allow
xinclude processing (i.e. either MAY or MUST) then the result of a GRDDL
transform is non-deterministic - it depends on whether the XINCLUDEd
document can be loaded.

I don't know if this has been mentioned before, but that rather than us
trying to give a definitive answer on xinclude processing, we could
leave it to the publisher.  We could allow the publisher to specify
whether to do xinclude processing before running the transform, e.g. we
have "transform" and "transformall" indicators for transforms to apply.

If we allow XINCLUDE processing, then other questions arise.  What if
the included document has a GRDDL transform specified on its root node,
or in some other way?  

Brian

[1] http://www.w3.org/TR/xpath20/
[2] http://www.w3.org/TR/xinclude/

Received on Wednesday, 13 December 2006 09:17:16 UTC