I'd even argue for a minimum *without attributes* and without mixed content,
no DTD subset, no namespaces, and only UTF8 support.
A processor (parser/ language binding etc) for such a minimum would be much
smaller, possibly "JavaScript" small/fast,
and on equal footing with JSON.
It would still be "XML" (i.e. parseable by full parsers) but vastly useful
on its own.
I personally don't see the point in this. If I don't need mixed content or attributes, I use JSON.
I think we need to focus on the greatest need. I considered responding to the earlier "hysteria" comment by saying that most folks on this thread, and certainly James Clark's post were not "running scared" of JSON. For my part, I embrace it, and I know quite a few others do who still want XML to be simplified.
That makes me think that the biggest sweet spot is the minimal subset of XML that provides real supplementary benefits over JSON, which is:
no namespaces (or at least radically simplified NS), no DTD and only UTF-8, UTF-16 or UTF-32 (I disagree with UTF-8 only, and I think you might have some trouble with that in some geographies, even if you could get everyone to accept Unicode only in the first place). I'm meh on tinkering with comments or PIs, and I'm 100% against any monolitihic datatyping, though I'd be up for a *simple* but extensible annotation mechanism.
