Re: [xml-dev] Re: determining ID-ness in XML

[Rick Jelliffe <ricko@allette.com.au>]

 From: "Henry S. Thompson" <ht@cogsci.ed.ac.uk>
 
> 2) They're not declared in the grammar (DTD/Schema), so their usage is not
> subject to any declarative/universal quality control: you're back to
> writing ad-hoc code in every application to check they occur where and 
> how they're supposed to.

But they are not subject to declarations because schema specification 
committees have decided not cover them.  

"Schema languages should not support PIs because people don't use them; 
people should not use PIs because schema languages don't support them."

Looks circular. 

And, in any case, in SGML you can declare an attribute is a PI as a type. 
The issue for PIs is not the <?brackets?>, but whether it is desirable to label 
information items as containing process-specific code (i.e. PIs) and
non-XML code (notations). 

Cheer
Rick Jelliffe