Lists Home |
Date Index |
From: "James Clark" <firstname.lastname@example.org>
> I think Erik's right that attributes in XML are a mess. Either
> - they ought to have be able to structured values, just like elements (i.e.
> one could view the content of an element as a special unnamed attribute), or
> - they should be removed entirely.
Is there any structured "attribute" which cannot be represented now using IDREF?
Why invent a new syntax, or a new naming convention (xml:information-item="attribute"?),
or a new pragma annotator?
XML has two pragmatics built-in that almost everyone uses:
- some children of elements can be relied on to be available unordered
before the other children are, to help stream processing (attributes)
- some children of elements are global, unconstrained and, in a pinch, forgettable
(PIs and comments)
It may be that RELAX NG's interleave (or Schematron too) is powerful
enough that a schema language can express this constraint. But why make
it unreliable: it needs to be a syntactic (or architectural thing) rather than
a schema constraint.
> The non-uniformity between elements and attributes has a huge cost.
But how much is that cost compared to the costs of "improvement".