RE: PI for associating a sample input with a styleshee

> I am curious how much interest is in having such a PI. I 
> remember that I was pro having a PI to associate a Relax NG 
> schema with an XML document and there was a strong reaction 
> against making a proposal for such a PI on the Relax NG list. 
> So what people think about such a PI?

I think the problem is that a PI is an "instruction", it is imperative. I
don't believe that a document should contain instructions that say how it is
to be processed. That belongs elsewhere, in a task or workload or script or
scenario...

However, I think there's room for assertions. It's reasonable for an
instance document to contain an assertion that it conforms to a particular
schema. That implies multiple such assertions should be allowed. It's also
reasonable for a stylesheet to contain assertions about the types of its
input and output.

But in an IDE used for XML/XSLT transformations, I would like to be able to
select from a list of tasks of the form (transform X using Y with parameters
P and processor Z) where there are no restrictions on having the same X or Y
in as many tasks as you like. I certainly don't like having the IDE modify
my source document when I run such a task.

Michael Kay
http://www.saxonica.com/