A quick update on our discussion of PEGs and ordered choice after a bit more reading. 

The way PEG implements ordered choice appears to reduce the language recognized by some grammars, rather than just reducing the number of parse trees produced in the case of an ambiguity. (This is in comparison with the same grammar using unordered choice.)

I can't persuade myself that ordered choice will do this in all cases, though, and I've found a couple of papers proposing an ordered-choice context-free formalism. So it might be possible to do ordered choice while keeping iXML context-free. 

I'll try to finish up the document about approaches to ambiguity resolution next week. I think we need to discuss whether we have any interest in extending iXML beyond context-free or not, because that will really affect which kinds of enhancements we're willing to consider and which we aren't. 

Best,

BTW