Re: Thoughts on pragmas and iXML

Bethan Tovey-Walsh <bytheway@linguacelta.com> writes:
> I suspect that Requirement 9 is one that doesn't conform to the PI model, since it allows pragmas to attach to more than one construct.

I think it applies as well as it can. XML and iXML have different abstract models to content with, but PIs can definitely go in many well-defined places:

<?pi1 before the root element?>
<document-element>
  <?pi2 in an element?>
  <child>
    <?pi3 here too?>
  </child>
  <?p4 between elements?>
  <child/>
</document-element>
<?pi5 after the root?>

In practice, all of the PIs that I’m aware of that apply to the document as a whole have to go before the document element. Within a document, I think the prevailing interpretation is that a PI applies to its parent, but you could describe a PI that applied to it’s following sibling, or some other place. “Applies to my great aunt’s third niece” doesn’t seem like a very convenient way to specify behavior but it could be done.

The same is true in iXML, of course. I think we should describe how pragmas are placed into the XML serialization, and I think we should do it in a way that’s consistent and predictable. But that doesn’t constrain the implementor in any way.

Having them scattered about haphazardly, the way the current grammar places comments into the XML serialization, is just aggressively hostile to users, IMHO.

                                        Be seeing you,
                                          norm

--
Norm Tovey-Walsh
Saxonica

Received on Monday, 17 February 2025 19:57:21 UTC